Line data Source code
1 : use std::collections::{HashMap, HashSet};
2 : use std::path::PathBuf;
3 : use std::str::FromStr;
4 : use std::time::Duration;
5 :
6 : use clap::{Parser, Subcommand};
7 : use futures::StreamExt;
8 : use pageserver_api::controller_api::{
9 : AvailabilityZone, MigrationConfig, NodeAvailabilityWrapper, NodeConfigureRequest,
10 : NodeDescribeResponse, NodeRegisterRequest, NodeSchedulingPolicy, NodeShardResponse,
11 : PlacementPolicy, SafekeeperDescribeResponse, SafekeeperSchedulingPolicyRequest,
12 : ShardSchedulingPolicy, ShardsPreferredAzsRequest, ShardsPreferredAzsResponse,
13 : SkSchedulingPolicy, TenantCreateRequest, TenantDescribeResponse, TenantPolicyRequest,
14 : TenantShardMigrateRequest, TenantShardMigrateResponse, TimelineSafekeeperMigrateRequest,
15 : };
16 : use pageserver_api::models::{
17 : EvictionPolicy, EvictionPolicyLayerAccessThreshold, ShardParameters, TenantConfig,
18 : TenantConfigPatchRequest, TenantConfigRequest, TenantShardSplitRequest,
19 : TenantShardSplitResponse,
20 : };
21 : use pageserver_api::shard::{ShardStripeSize, TenantShardId};
22 : use pageserver_client::mgmt_api::{self};
23 : use reqwest::{Certificate, Method, StatusCode, Url};
24 : use safekeeper_api::models::TimelineLocateResponse;
25 : use storage_controller_client::control_api::Client;
26 : use utils::id::{NodeId, TenantId, TimelineId};
27 :
28 : #[derive(Subcommand, Debug)]
29 : enum Command {
30 : /// Register a pageserver with the storage controller. This shouldn't usually be necessary,
31 : /// since pageservers auto-register when they start up
32 : NodeRegister {
33 : #[arg(long)]
34 : node_id: NodeId,
35 :
36 : #[arg(long)]
37 : listen_pg_addr: String,
38 : #[arg(long)]
39 : listen_pg_port: u16,
40 : #[arg(long)]
41 : listen_grpc_addr: Option<String>,
42 : #[arg(long)]
43 : listen_grpc_port: Option<u16>,
44 :
45 : #[arg(long)]
46 : listen_http_addr: String,
47 : #[arg(long)]
48 : listen_http_port: u16,
49 : #[arg(long)]
50 : listen_https_port: Option<u16>,
51 :
52 : #[arg(long)]
53 : availability_zone_id: String,
54 : },
55 :
56 : /// Modify a node's configuration in the storage controller
57 : NodeConfigure {
58 : #[arg(long)]
59 : node_id: NodeId,
60 :
61 : /// Availability is usually auto-detected based on heartbeats. Set 'offline' here to
62 : /// manually mark a node offline
63 : #[arg(long)]
64 : availability: Option<NodeAvailabilityArg>,
65 : /// Scheduling policy controls whether tenant shards may be scheduled onto this node.
66 : #[arg(long)]
67 : scheduling: Option<NodeSchedulingPolicy>,
68 : },
69 : /// Exists for backup usage and will be removed in future.
70 : /// Use [`Command::NodeStartDelete`] instead, if possible.
71 : NodeDelete {
72 : #[arg(long)]
73 : node_id: NodeId,
74 : },
75 : /// Start deletion of the specified pageserver.
76 : NodeStartDelete {
77 : #[arg(long)]
78 : node_id: NodeId,
79 : },
80 : /// Cancel deletion of the specified pageserver and wait for `timeout`
81 : /// for the operation to be canceled. May be retried.
82 : NodeCancelDelete {
83 : #[arg(long)]
84 : node_id: NodeId,
85 : #[arg(long)]
86 : timeout: humantime::Duration,
87 : },
88 : /// Delete a tombstone of node from the storage controller.
89 : /// This is used when we want to allow the node to be re-registered.
90 : NodeDeleteTombstone {
91 : #[arg(long)]
92 : node_id: NodeId,
93 : },
94 : /// Modify a tenant's policies in the storage controller
95 : TenantPolicy {
96 : #[arg(long)]
97 : tenant_id: TenantId,
98 : /// Placement policy controls whether a tenant is `detached`, has only a secondary location (`secondary`),
99 : /// or is in the normal attached state with N secondary locations (`attached:N`)
100 : #[arg(long)]
101 : placement: Option<PlacementPolicyArg>,
102 : /// Scheduling policy enables pausing the controller's scheduling activity involving this tenant. `active` is normal,
103 : /// `essential` disables optimization scheduling changes, `pause` disables all scheduling changes, and `stop` prevents
104 : /// all reconciliation activity including for scheduling changes already made. `pause` and `stop` can make a tenant
105 : /// unavailable, and are only for use in emergencies.
106 : #[arg(long)]
107 : scheduling: Option<ShardSchedulingPolicyArg>,
108 : },
109 : /// List nodes known to the storage controller
110 : Nodes {},
111 : /// List soft deleted nodes known to the storage controller
112 : NodeTombstones {},
113 : /// List tenants known to the storage controller
114 : Tenants {
115 : /// If this field is set, it will list the tenants on a specific node
116 : node_id: Option<NodeId>,
117 : },
118 : /// Create a new tenant in the storage controller, and by extension on pageservers.
119 : TenantCreate {
120 : #[arg(long)]
121 : tenant_id: TenantId,
122 : },
123 : /// Delete a tenant in the storage controller, and by extension on pageservers.
124 : TenantDelete {
125 : #[arg(long)]
126 : tenant_id: TenantId,
127 : },
128 : /// Split an existing tenant into a higher number of shards than its current shard count.
129 : TenantShardSplit {
130 : #[arg(long)]
131 : tenant_id: TenantId,
132 : #[arg(long)]
133 : shard_count: u8,
134 : /// Optional, in 8kiB pages. e.g. set 2048 for 16MB stripes.
135 : #[arg(long)]
136 : stripe_size: Option<u32>,
137 : },
138 : /// Migrate the attached location for a tenant shard to a specific pageserver.
139 : TenantShardMigrate {
140 : #[arg(long)]
141 : tenant_shard_id: TenantShardId,
142 : #[arg(long)]
143 : node: NodeId,
144 : #[arg(long, default_value_t = true, action = clap::ArgAction::Set)]
145 : prewarm: bool,
146 : #[arg(long, default_value_t = false, action = clap::ArgAction::Set)]
147 : override_scheduler: bool,
148 : },
149 : /// Watch the location of a tenant shard evolve, e.g. while expecting it to migrate
150 : TenantShardWatch {
151 : #[arg(long)]
152 : tenant_shard_id: TenantShardId,
153 : },
154 : /// Migrate the secondary location for a tenant shard to a specific pageserver.
155 : TenantShardMigrateSecondary {
156 : #[arg(long)]
157 : tenant_shard_id: TenantShardId,
158 : #[arg(long)]
159 : node: NodeId,
160 : },
161 : /// Cancel any ongoing reconciliation for this shard
162 : TenantShardCancelReconcile {
163 : #[arg(long)]
164 : tenant_shard_id: TenantShardId,
165 : },
166 : /// Set the pageserver tenant configuration of a tenant: this is the configuration structure
167 : /// that is passed through to pageservers, and does not affect storage controller behavior.
168 : /// Any previous tenant configs are overwritten.
169 : SetTenantConfig {
170 : #[arg(long)]
171 : tenant_id: TenantId,
172 : #[arg(long)]
173 : config: String,
174 : },
175 : /// Patch the pageserver tenant configuration of a tenant. Any fields with null values in the
176 : /// provided JSON are unset from the tenant config and all fields with non-null values are set.
177 : /// Unspecified fields are not changed.
178 : PatchTenantConfig {
179 : #[arg(long)]
180 : tenant_id: TenantId,
181 : #[arg(long)]
182 : config: String,
183 : },
184 : /// Print details about a particular tenant, including all its shards' states.
185 : TenantDescribe {
186 : #[arg(long)]
187 : tenant_id: TenantId,
188 : },
189 : TenantSetPreferredAz {
190 : #[arg(long)]
191 : tenant_id: TenantId,
192 : #[arg(long)]
193 : preferred_az: Option<String>,
194 : },
195 : /// Uncleanly drop a tenant from the storage controller: this doesn't delete anything from pageservers. Appropriate
196 : /// if you e.g. used `tenant-warmup` by mistake on a tenant ID that doesn't really exist, or is in some other region.
197 : TenantDrop {
198 : #[arg(long)]
199 : tenant_id: TenantId,
200 : #[arg(long)]
201 : unclean: bool,
202 : },
203 : NodeDrop {
204 : #[arg(long)]
205 : node_id: NodeId,
206 : #[arg(long)]
207 : unclean: bool,
208 : },
209 : TenantSetTimeBasedEviction {
210 : #[arg(long)]
211 : tenant_id: TenantId,
212 : #[arg(long)]
213 : period: humantime::Duration,
214 : #[arg(long)]
215 : threshold: humantime::Duration,
216 : },
217 : // Migrate away from a set of specified pageservers by moving the primary attachments to pageservers
218 : // outside of the specified set.
219 : BulkMigrate {
220 : // Set of pageserver node ids to drain.
221 : #[arg(long)]
222 : nodes: Vec<NodeId>,
223 : // Optional: migration concurrency (default is 8)
224 : #[arg(long)]
225 : concurrency: Option<usize>,
226 : // Optional: maximum number of shards to migrate
227 : #[arg(long)]
228 : max_shards: Option<usize>,
229 : // Optional: when set to true, nothing is migrated, but the plan is printed to stdout
230 : #[arg(long)]
231 : dry_run: Option<bool>,
232 : },
233 : /// Start draining the specified pageserver.
234 : /// The drain is complete when the schedulling policy returns to active.
235 : StartDrain {
236 : #[arg(long)]
237 : node_id: NodeId,
238 : },
239 : /// Cancel draining the specified pageserver and wait for `timeout`
240 : /// for the operation to be canceled. May be retried.
241 : CancelDrain {
242 : #[arg(long)]
243 : node_id: NodeId,
244 : #[arg(long)]
245 : timeout: humantime::Duration,
246 : },
247 : /// Start filling the specified pageserver.
248 : /// The drain is complete when the schedulling policy returns to active.
249 : StartFill {
250 : #[arg(long)]
251 : node_id: NodeId,
252 : },
253 : /// Cancel filling the specified pageserver and wait for `timeout`
254 : /// for the operation to be canceled. May be retried.
255 : CancelFill {
256 : #[arg(long)]
257 : node_id: NodeId,
258 : #[arg(long)]
259 : timeout: humantime::Duration,
260 : },
261 : /// List safekeepers known to the storage controller
262 : Safekeepers {},
263 : /// Set the scheduling policy of the specified safekeeper
264 : SafekeeperScheduling {
265 : #[arg(long)]
266 : node_id: NodeId,
267 : #[arg(long)]
268 : scheduling_policy: SkSchedulingPolicyArg,
269 : },
270 : /// Downloads any missing heatmap layers for all shard for a given timeline
271 : DownloadHeatmapLayers {
272 : /// Tenant ID or tenant shard ID. When an unsharded tenant ID is specified,
273 : /// the operation is performed on all shards. When a sharded tenant ID is
274 : /// specified, the operation is only performed on the specified shard.
275 : #[arg(long)]
276 : tenant_shard_id: TenantShardId,
277 : #[arg(long)]
278 : timeline_id: TimelineId,
279 : /// Optional: Maximum download concurrency (default is 16)
280 : #[arg(long)]
281 : concurrency: Option<usize>,
282 : },
283 : /// Locate safekeepers for a timeline from the storcon DB.
284 : TimelineLocate {
285 : #[arg(long)]
286 : tenant_id: TenantId,
287 : #[arg(long)]
288 : timeline_id: TimelineId,
289 : },
290 : /// Migrate a timeline to a new set of safekeepers
291 : TimelineSafekeeperMigrate {
292 : #[arg(long)]
293 : tenant_id: TenantId,
294 : #[arg(long)]
295 : timeline_id: TimelineId,
296 : /// Example: --new-sk-set 1,2,3
297 : #[arg(long, required = true, value_delimiter = ',')]
298 : new_sk_set: Vec<NodeId>,
299 : },
300 : }
301 :
302 : #[derive(Parser)]
303 : #[command(
304 : author,
305 : version,
306 : about,
307 : long_about = "CLI for Storage Controller Support/Debug"
308 : )]
309 : #[command(arg_required_else_help(true))]
310 : struct Cli {
311 : #[arg(long)]
312 : /// URL to storage controller. e.g. http://127.0.0.1:1234 when using `neon_local`
313 : api: Url,
314 :
315 : #[arg(long)]
316 : /// JWT token for authenticating with storage controller. Depending on the API used, this
317 : /// should have either `pageserverapi` or `admin` scopes: for convenience, you should mint
318 : /// a token with both scopes to use with this tool.
319 : jwt: Option<String>,
320 :
321 : #[arg(long)]
322 : /// Trusted root CA certificates to use in https APIs.
323 : ssl_ca_file: Option<PathBuf>,
324 :
325 : #[command(subcommand)]
326 : command: Command,
327 : }
328 :
329 : #[derive(Debug, Clone)]
330 : struct PlacementPolicyArg(PlacementPolicy);
331 :
332 : impl FromStr for PlacementPolicyArg {
333 : type Err = anyhow::Error;
334 :
335 0 : fn from_str(s: &str) -> Result<Self, Self::Err> {
336 0 : match s {
337 0 : "detached" => Ok(Self(PlacementPolicy::Detached)),
338 0 : "secondary" => Ok(Self(PlacementPolicy::Secondary)),
339 0 : _ if s.starts_with("attached:") => {
340 0 : let mut splitter = s.split(':');
341 0 : let _prefix = splitter.next().unwrap();
342 0 : match splitter.next().and_then(|s| s.parse::<usize>().ok()) {
343 0 : Some(n) => Ok(Self(PlacementPolicy::Attached(n))),
344 0 : None => Err(anyhow::anyhow!(
345 0 : "Invalid format '{s}', a valid example is 'attached:1'"
346 0 : )),
347 : }
348 : }
349 0 : _ => Err(anyhow::anyhow!(
350 0 : "Unknown placement policy '{s}', try detached,secondary,attached:<n>"
351 0 : )),
352 : }
353 0 : }
354 : }
355 :
356 : #[derive(Debug, Clone)]
357 : struct SkSchedulingPolicyArg(SkSchedulingPolicy);
358 :
359 : impl FromStr for SkSchedulingPolicyArg {
360 : type Err = anyhow::Error;
361 :
362 0 : fn from_str(s: &str) -> Result<Self, Self::Err> {
363 0 : SkSchedulingPolicy::from_str(s).map(Self)
364 0 : }
365 : }
366 :
367 : #[derive(Debug, Clone)]
368 : struct ShardSchedulingPolicyArg(ShardSchedulingPolicy);
369 :
370 : impl FromStr for ShardSchedulingPolicyArg {
371 : type Err = anyhow::Error;
372 :
373 0 : fn from_str(s: &str) -> Result<Self, Self::Err> {
374 0 : match s {
375 0 : "active" => Ok(Self(ShardSchedulingPolicy::Active)),
376 0 : "essential" => Ok(Self(ShardSchedulingPolicy::Essential)),
377 0 : "pause" => Ok(Self(ShardSchedulingPolicy::Pause)),
378 0 : "stop" => Ok(Self(ShardSchedulingPolicy::Stop)),
379 0 : _ => Err(anyhow::anyhow!(
380 0 : "Unknown scheduling policy '{s}', try active,essential,pause,stop"
381 0 : )),
382 : }
383 0 : }
384 : }
385 :
386 : #[derive(Debug, Clone)]
387 : struct NodeAvailabilityArg(NodeAvailabilityWrapper);
388 :
389 : impl FromStr for NodeAvailabilityArg {
390 : type Err = anyhow::Error;
391 :
392 0 : fn from_str(s: &str) -> Result<Self, Self::Err> {
393 0 : match s {
394 0 : "active" => Ok(Self(NodeAvailabilityWrapper::Active)),
395 0 : "offline" => Ok(Self(NodeAvailabilityWrapper::Offline)),
396 0 : _ => Err(anyhow::anyhow!("Unknown availability state '{s}'")),
397 : }
398 0 : }
399 : }
400 :
401 0 : async fn wait_for_scheduling_policy<F>(
402 0 : client: Client,
403 0 : node_id: NodeId,
404 0 : timeout: Duration,
405 0 : f: F,
406 0 : ) -> anyhow::Result<NodeSchedulingPolicy>
407 0 : where
408 0 : F: Fn(NodeSchedulingPolicy) -> bool,
409 0 : {
410 0 : let waiter = tokio::time::timeout(timeout, async move {
411 : loop {
412 0 : let node = client
413 0 : .dispatch::<(), NodeDescribeResponse>(
414 0 : Method::GET,
415 0 : format!("control/v1/node/{node_id}"),
416 0 : None,
417 0 : )
418 0 : .await?;
419 :
420 0 : if f(node.scheduling) {
421 0 : return Ok::<NodeSchedulingPolicy, mgmt_api::Error>(node.scheduling);
422 0 : }
423 : }
424 0 : });
425 :
426 0 : Ok(waiter.await??)
427 0 : }
428 :
429 : #[tokio::main]
430 0 : async fn main() -> anyhow::Result<()> {
431 0 : let cli = Cli::parse();
432 :
433 0 : let ssl_ca_certs = match &cli.ssl_ca_file {
434 0 : Some(ssl_ca_file) => {
435 0 : let buf = tokio::fs::read(ssl_ca_file).await?;
436 0 : Certificate::from_pem_bundle(&buf)?
437 : }
438 0 : None => Vec::new(),
439 : };
440 :
441 0 : let mut http_client = reqwest::Client::builder();
442 0 : for ssl_ca_cert in ssl_ca_certs {
443 0 : http_client = http_client.add_root_certificate(ssl_ca_cert);
444 0 : }
445 0 : let http_client = http_client.build()?;
446 :
447 0 : let storcon_client = Client::new(http_client.clone(), cli.api.clone(), cli.jwt.clone());
448 :
449 0 : let mut trimmed = cli.api.to_string();
450 0 : trimmed.pop();
451 0 : let vps_client = mgmt_api::Client::new(http_client.clone(), trimmed, cli.jwt.as_deref());
452 :
453 0 : match cli.command {
454 0 : Command::NodeRegister {
455 0 : node_id,
456 0 : listen_pg_addr,
457 0 : listen_pg_port,
458 0 : listen_grpc_addr,
459 0 : listen_grpc_port,
460 0 : listen_http_addr,
461 0 : listen_http_port,
462 0 : listen_https_port,
463 0 : availability_zone_id,
464 0 : } => {
465 0 : storcon_client
466 0 : .dispatch::<_, ()>(
467 0 : Method::POST,
468 0 : "control/v1/node".to_string(),
469 0 : Some(NodeRegisterRequest {
470 0 : node_id,
471 0 : listen_pg_addr,
472 0 : listen_pg_port,
473 0 : listen_grpc_addr,
474 0 : listen_grpc_port,
475 0 : listen_http_addr,
476 0 : listen_http_port,
477 0 : listen_https_port,
478 0 : availability_zone_id: AvailabilityZone(availability_zone_id),
479 0 : node_ip_addr: None,
480 0 : }),
481 0 : )
482 0 : .await?;
483 0 : }
484 0 : Command::TenantCreate { tenant_id } => {
485 0 : storcon_client
486 0 : .dispatch::<_, ()>(
487 0 : Method::POST,
488 0 : "v1/tenant".to_string(),
489 0 : Some(TenantCreateRequest {
490 0 : new_tenant_id: TenantShardId::unsharded(tenant_id),
491 0 : generation: None,
492 0 : shard_parameters: ShardParameters::default(),
493 0 : placement_policy: Some(PlacementPolicy::Attached(1)),
494 0 : config: TenantConfig::default(),
495 0 : }),
496 0 : )
497 0 : .await?;
498 0 : }
499 0 : Command::TenantDelete { tenant_id } => {
500 0 : let status = vps_client
501 0 : .tenant_delete(TenantShardId::unsharded(tenant_id))
502 0 : .await?;
503 0 : tracing::info!("Delete status: {}", status);
504 0 : }
505 0 : Command::Nodes {} => {
506 0 : let mut resp = storcon_client
507 0 : .dispatch::<(), Vec<NodeDescribeResponse>>(
508 0 : Method::GET,
509 0 : "control/v1/node".to_string(),
510 0 : None,
511 0 : )
512 0 : .await?;
513 0 :
514 0 : resp.sort_by(|a, b| a.listen_http_addr.cmp(&b.listen_http_addr));
515 0 :
516 0 : let mut table = comfy_table::Table::new();
517 0 : table.set_header(["Id", "Hostname", "AZ", "Scheduling", "Availability"]);
518 0 : for node in resp {
519 0 : table.add_row([
520 0 : format!("{}", node.id),
521 0 : node.listen_http_addr,
522 0 : node.availability_zone_id,
523 0 : format!("{:?}", node.scheduling),
524 0 : format!("{:?}", node.availability),
525 0 : ]);
526 0 : }
527 0 : println!("{table}");
528 0 : }
529 0 : Command::NodeConfigure {
530 0 : node_id,
531 0 : availability,
532 0 : scheduling,
533 0 : } => {
534 0 : let req = NodeConfigureRequest {
535 0 : node_id,
536 0 : availability: availability.map(|a| a.0),
537 0 : scheduling,
538 0 : };
539 0 : storcon_client
540 0 : .dispatch::<_, ()>(
541 0 : Method::PUT,
542 0 : format!("control/v1/node/{node_id}/config"),
543 0 : Some(req),
544 0 : )
545 0 : .await?;
546 0 : }
547 0 : Command::Tenants {
548 0 : node_id: Some(node_id),
549 0 : } => {
550 0 : let describe_response = storcon_client
551 0 : .dispatch::<(), NodeShardResponse>(
552 0 : Method::GET,
553 0 : format!("control/v1/node/{node_id}/shards"),
554 0 : None,
555 0 : )
556 0 : .await?;
557 0 : let shards = describe_response.shards;
558 0 : let mut table = comfy_table::Table::new();
559 0 : table.set_header([
560 0 : "Shard",
561 0 : "Intended Primary/Secondary",
562 0 : "Observed Primary/Secondary",
563 0 : ]);
564 0 : for shard in shards {
565 0 : table.add_row([
566 0 : format!("{}", shard.tenant_shard_id),
567 0 : match shard.is_intended_secondary {
568 0 : None => "".to_string(),
569 0 : Some(true) => "Secondary".to_string(),
570 0 : Some(false) => "Primary".to_string(),
571 0 : },
572 0 : match shard.is_observed_secondary {
573 0 : None => "".to_string(),
574 0 : Some(true) => "Secondary".to_string(),
575 0 : Some(false) => "Primary".to_string(),
576 0 : },
577 0 : ]);
578 0 : }
579 0 : println!("{table}");
580 0 : }
581 0 : Command::Tenants { node_id: None } => {
582 0 : // Set up output formatting
583 0 : let mut table = comfy_table::Table::new();
584 0 : table.set_header([
585 0 : "TenantId",
586 0 : "Preferred AZ",
587 0 : "ShardCount",
588 0 : "StripeSize",
589 0 : "Placement",
590 0 : "Scheduling",
591 0 : ]);
592 0 :
593 0 : // Pagination loop over listing API
594 0 : let mut start_after = None;
595 0 : const LIMIT: usize = 1000;
596 0 : loop {
597 0 : let path = match start_after {
598 0 : None => format!("control/v1/tenant?limit={LIMIT}"),
599 0 : Some(start_after) => {
600 0 : format!("control/v1/tenant?limit={LIMIT}&start_after={start_after}")
601 0 : }
602 0 : };
603 0 :
604 0 : let resp = storcon_client
605 0 : .dispatch::<(), Vec<TenantDescribeResponse>>(Method::GET, path, None)
606 0 : .await?;
607 0 :
608 0 : if resp.is_empty() {
609 0 : // End of data reached
610 0 : break;
611 0 : }
612 0 :
613 0 : // Give some visual feedback while we're building up the table (comfy_table doesn't have
614 0 : // streaming output)
615 0 : if resp.len() >= LIMIT {
616 0 : eprint!(".");
617 0 : }
618 0 :
619 0 : start_after = Some(resp.last().unwrap().tenant_id);
620 0 :
621 0 : for tenant in resp {
622 0 : let shard_zero = tenant.shards.into_iter().next().unwrap();
623 0 : table.add_row([
624 0 : format!("{}", tenant.tenant_id),
625 0 : shard_zero
626 0 : .preferred_az_id
627 0 : .as_ref()
628 0 : .cloned()
629 0 : .unwrap_or("".to_string()),
630 0 : format!("{}", shard_zero.tenant_shard_id.shard_count.literal()),
631 0 : format!("{:?}", tenant.stripe_size),
632 0 : format!("{:?}", tenant.policy),
633 0 : format!("{:?}", shard_zero.scheduling_policy),
634 0 : ]);
635 0 : }
636 0 : }
637 0 :
638 0 : // Terminate progress dots
639 0 : if table.row_count() > LIMIT {
640 0 : eprint!("");
641 0 : }
642 0 :
643 0 : println!("{table}");
644 0 : }
645 0 : Command::TenantPolicy {
646 0 : tenant_id,
647 0 : placement,
648 0 : scheduling,
649 0 : } => {
650 0 : let req = TenantPolicyRequest {
651 0 : scheduling: scheduling.map(|s| s.0),
652 0 : placement: placement.map(|p| p.0),
653 0 : };
654 0 : storcon_client
655 0 : .dispatch::<_, ()>(
656 0 : Method::PUT,
657 0 : format!("control/v1/tenant/{tenant_id}/policy"),
658 0 : Some(req),
659 0 : )
660 0 : .await?;
661 0 : }
662 0 : Command::TenantShardSplit {
663 0 : tenant_id,
664 0 : shard_count,
665 0 : stripe_size,
666 0 : } => {
667 0 : let req = TenantShardSplitRequest {
668 0 : new_shard_count: shard_count,
669 0 : new_stripe_size: stripe_size.map(ShardStripeSize),
670 0 : };
671 0 :
672 0 : let response = storcon_client
673 0 : .dispatch::<TenantShardSplitRequest, TenantShardSplitResponse>(
674 0 : Method::PUT,
675 0 : format!("control/v1/tenant/{tenant_id}/shard_split"),
676 0 : Some(req),
677 0 : )
678 0 : .await?;
679 0 : println!(
680 0 : "Split tenant {} into {} shards: {}",
681 0 : tenant_id,
682 0 : shard_count,
683 0 : response
684 0 : .new_shards
685 0 : .iter()
686 0 : .map(|s| format!("{s:?}"))
687 0 : .collect::<Vec<_>>()
688 0 : .join(",")
689 0 : );
690 0 : }
691 0 : Command::TenantShardMigrate {
692 0 : tenant_shard_id,
693 0 : node,
694 0 : prewarm,
695 0 : override_scheduler,
696 0 : } => {
697 0 : let migration_config = MigrationConfig {
698 0 : prewarm,
699 0 : override_scheduler,
700 0 : ..Default::default()
701 0 : };
702 0 :
703 0 : let req = TenantShardMigrateRequest {
704 0 : node_id: node,
705 0 : origin_node_id: None,
706 0 : migration_config,
707 0 : };
708 0 :
709 0 : match storcon_client
710 0 : .dispatch::<TenantShardMigrateRequest, TenantShardMigrateResponse>(
711 0 : Method::PUT,
712 0 : format!("control/v1/tenant/{tenant_shard_id}/migrate"),
713 0 : Some(req),
714 0 : )
715 0 : .await
716 0 : {
717 0 : Err(mgmt_api::Error::ApiError(StatusCode::PRECONDITION_FAILED, msg)) => {
718 0 : anyhow::bail!(
719 0 : "Migration to {node} rejected, may require `--force` ({}) ",
720 0 : msg
721 0 : );
722 0 : }
723 0 : Err(e) => return Err(e.into()),
724 0 : Ok(_) => {}
725 0 : }
726 0 :
727 0 : watch_tenant_shard(storcon_client, tenant_shard_id, Some(node)).await?;
728 0 : }
729 0 : Command::TenantShardWatch { tenant_shard_id } => {
730 0 : watch_tenant_shard(storcon_client, tenant_shard_id, None).await?;
731 0 : }
732 0 : Command::TenantShardMigrateSecondary {
733 0 : tenant_shard_id,
734 0 : node,
735 0 : } => {
736 0 : let req = TenantShardMigrateRequest {
737 0 : node_id: node,
738 0 : origin_node_id: None,
739 0 : migration_config: MigrationConfig::default(),
740 0 : };
741 0 :
742 0 : storcon_client
743 0 : .dispatch::<TenantShardMigrateRequest, TenantShardMigrateResponse>(
744 0 : Method::PUT,
745 0 : format!("control/v1/tenant/{tenant_shard_id}/migrate_secondary"),
746 0 : Some(req),
747 0 : )
748 0 : .await?;
749 0 : }
750 0 : Command::TenantShardCancelReconcile { tenant_shard_id } => {
751 0 : storcon_client
752 0 : .dispatch::<(), ()>(
753 0 : Method::PUT,
754 0 : format!("control/v1/tenant/{tenant_shard_id}/cancel_reconcile"),
755 0 : None,
756 0 : )
757 0 : .await?;
758 0 : }
759 0 : Command::SetTenantConfig { tenant_id, config } => {
760 0 : let tenant_conf = serde_json::from_str(&config)?;
761 0 :
762 0 : vps_client
763 0 : .set_tenant_config(&TenantConfigRequest {
764 0 : tenant_id,
765 0 : config: tenant_conf,
766 0 : })
767 0 : .await?;
768 0 : }
769 0 : Command::PatchTenantConfig { tenant_id, config } => {
770 0 : let tenant_conf = serde_json::from_str(&config)?;
771 0 :
772 0 : vps_client
773 0 : .patch_tenant_config(&TenantConfigPatchRequest {
774 0 : tenant_id,
775 0 : config: tenant_conf,
776 0 : })
777 0 : .await?;
778 0 : }
779 0 : Command::TenantDescribe { tenant_id } => {
780 0 : let TenantDescribeResponse {
781 0 : tenant_id,
782 0 : shards,
783 0 : stripe_size,
784 0 : policy,
785 0 : config,
786 0 : } = storcon_client
787 0 : .dispatch::<(), TenantDescribeResponse>(
788 0 : Method::GET,
789 0 : format!("control/v1/tenant/{tenant_id}"),
790 0 : None,
791 0 : )
792 0 : .await?;
793 0 :
794 0 : let nodes = storcon_client
795 0 : .dispatch::<(), Vec<NodeDescribeResponse>>(
796 0 : Method::GET,
797 0 : "control/v1/node".to_string(),
798 0 : None,
799 0 : )
800 0 : .await?;
801 0 : let nodes = nodes
802 0 : .into_iter()
803 0 : .map(|n| (n.id, n))
804 0 : .collect::<HashMap<_, _>>();
805 0 :
806 0 : println!("Tenant {tenant_id}");
807 0 : let mut table = comfy_table::Table::new();
808 0 : table.add_row(["Policy", &format!("{policy:?}")]);
809 0 : table.add_row(["Stripe size", &format!("{stripe_size:?}")]);
810 0 : table.add_row(["Config", &serde_json::to_string_pretty(&config).unwrap()]);
811 0 : println!("{table}");
812 0 : println!("Shards:");
813 0 : let mut table = comfy_table::Table::new();
814 0 : table.set_header([
815 0 : "Shard",
816 0 : "Attached",
817 0 : "Attached AZ",
818 0 : "Secondary",
819 0 : "Last error",
820 0 : "status",
821 0 : ]);
822 0 : for shard in shards {
823 0 : let secondary = shard
824 0 : .node_secondary
825 0 : .iter()
826 0 : .map(|n| format!("{n}"))
827 0 : .collect::<Vec<_>>()
828 0 : .join(",");
829 0 :
830 0 : let mut status_parts = Vec::new();
831 0 : if shard.is_reconciling {
832 0 : status_parts.push("reconciling");
833 0 : }
834 0 :
835 0 : if shard.is_pending_compute_notification {
836 0 : status_parts.push("pending_compute");
837 0 : }
838 0 :
839 0 : if shard.is_splitting {
840 0 : status_parts.push("splitting");
841 0 : }
842 0 : let status = status_parts.join(",");
843 0 :
844 0 : let attached_node = shard
845 0 : .node_attached
846 0 : .as_ref()
847 0 : .map(|id| nodes.get(id).expect("Shard references nonexistent node"));
848 0 :
849 0 : table.add_row([
850 0 : format!("{}", shard.tenant_shard_id),
851 0 : attached_node
852 0 : .map(|n| format!("{} ({})", n.listen_http_addr, n.id))
853 0 : .unwrap_or(String::new()),
854 0 : attached_node
855 0 : .map(|n| n.availability_zone_id.clone())
856 0 : .unwrap_or(String::new()),
857 0 : secondary,
858 0 : shard.last_error,
859 0 : status,
860 0 : ]);
861 0 : }
862 0 : println!("{table}");
863 0 : }
864 0 : Command::TenantSetPreferredAz {
865 0 : tenant_id,
866 0 : preferred_az,
867 0 : } => {
868 0 : // First learn about the tenant's shards
869 0 : let describe_response = storcon_client
870 0 : .dispatch::<(), TenantDescribeResponse>(
871 0 : Method::GET,
872 0 : format!("control/v1/tenant/{tenant_id}"),
873 0 : None,
874 0 : )
875 0 : .await?;
876 0 :
877 0 : // Learn about nodes to validate the AZ ID
878 0 : let nodes = storcon_client
879 0 : .dispatch::<(), Vec<NodeDescribeResponse>>(
880 0 : Method::GET,
881 0 : "control/v1/node".to_string(),
882 0 : None,
883 0 : )
884 0 : .await?;
885 0 :
886 0 : if let Some(preferred_az) = &preferred_az {
887 0 : let azs = nodes
888 0 : .into_iter()
889 0 : .map(|n| (n.availability_zone_id))
890 0 : .collect::<HashSet<_>>();
891 0 : if !azs.contains(preferred_az) {
892 0 : anyhow::bail!(
893 0 : "AZ {} not found on any node: known AZs are: {:?}",
894 0 : preferred_az,
895 0 : azs
896 0 : );
897 0 : }
898 0 : } else {
899 0 : // Make it obvious to the user that since they've omitted an AZ, we're clearing it
900 0 : eprintln!("Clearing preferred AZ for tenant {tenant_id}");
901 0 : }
902 0 :
903 0 : // Construct a request that modifies all the tenant's shards
904 0 : let req = ShardsPreferredAzsRequest {
905 0 : preferred_az_ids: describe_response
906 0 : .shards
907 0 : .into_iter()
908 0 : .map(|s| {
909 0 : (
910 0 : s.tenant_shard_id,
911 0 : preferred_az.clone().map(AvailabilityZone),
912 0 : )
913 0 : })
914 0 : .collect(),
915 0 : };
916 0 : storcon_client
917 0 : .dispatch::<ShardsPreferredAzsRequest, ShardsPreferredAzsResponse>(
918 0 : Method::PUT,
919 0 : "control/v1/preferred_azs".to_string(),
920 0 : Some(req),
921 0 : )
922 0 : .await?;
923 0 : }
924 0 : Command::TenantDrop { tenant_id, unclean } => {
925 0 : if !unclean {
926 0 : anyhow::bail!(
927 0 : "This command is not a tenant deletion, and uncleanly drops all controller state for the tenant. If you know what you're doing, add `--unclean` to proceed."
928 0 : )
929 0 : }
930 0 : storcon_client
931 0 : .dispatch::<(), ()>(
932 0 : Method::POST,
933 0 : format!("debug/v1/tenant/{tenant_id}/drop"),
934 0 : None,
935 0 : )
936 0 : .await?;
937 0 : }
938 0 : Command::NodeDrop { node_id, unclean } => {
939 0 : if !unclean {
940 0 : anyhow::bail!(
941 0 : "This command is not a clean node decommission, and uncleanly drops all controller state for the node, without checking if any tenants still refer to it. If you know what you're doing, add `--unclean` to proceed."
942 0 : )
943 0 : }
944 0 : storcon_client
945 0 : .dispatch::<(), ()>(Method::POST, format!("debug/v1/node/{node_id}/drop"), None)
946 0 : .await?;
947 0 : }
948 0 : Command::NodeDelete { node_id } => {
949 0 : eprintln!("Warning: This command is obsolete and will be removed in a future version");
950 0 : eprintln!("Use `NodeStartDelete` instead, if possible");
951 0 : storcon_client
952 0 : .dispatch::<(), ()>(Method::DELETE, format!("control/v1/node/{node_id}"), None)
953 0 : .await?;
954 0 : }
955 0 : Command::NodeStartDelete { node_id } => {
956 0 : storcon_client
957 0 : .dispatch::<(), ()>(
958 0 : Method::PUT,
959 0 : format!("control/v1/node/{node_id}/delete"),
960 0 : None,
961 0 : )
962 0 : .await?;
963 0 : println!("Delete started for {node_id}");
964 0 : }
965 0 : Command::NodeCancelDelete { node_id, timeout } => {
966 0 : storcon_client
967 0 : .dispatch::<(), ()>(
968 0 : Method::DELETE,
969 0 : format!("control/v1/node/{node_id}/delete"),
970 0 : None,
971 0 : )
972 0 : .await?;
973 0 :
974 0 : println!("Waiting for node {node_id} to quiesce on scheduling policy ...");
975 0 :
976 0 : let final_policy =
977 0 : wait_for_scheduling_policy(storcon_client, node_id, *timeout, |sched| {
978 0 : !matches!(sched, NodeSchedulingPolicy::Deleting)
979 0 : })
980 0 : .await?;
981 0 :
982 0 : println!(
983 0 : "Delete was cancelled for node {node_id}. Schedulling policy is now {final_policy:?}"
984 0 : );
985 0 : }
986 0 : Command::NodeDeleteTombstone { node_id } => {
987 0 : storcon_client
988 0 : .dispatch::<(), ()>(
989 0 : Method::DELETE,
990 0 : format!("debug/v1/tombstone/{node_id}"),
991 0 : None,
992 0 : )
993 0 : .await?;
994 0 : }
995 0 : Command::NodeTombstones {} => {
996 0 : let mut resp = storcon_client
997 0 : .dispatch::<(), Vec<NodeDescribeResponse>>(
998 0 : Method::GET,
999 0 : "debug/v1/tombstone".to_string(),
1000 0 : None,
1001 0 : )
1002 0 : .await?;
1003 0 :
1004 0 : resp.sort_by(|a, b| a.listen_http_addr.cmp(&b.listen_http_addr));
1005 0 :
1006 0 : let mut table = comfy_table::Table::new();
1007 0 : table.set_header(["Id", "Hostname", "AZ", "Scheduling", "Availability"]);
1008 0 : for node in resp {
1009 0 : table.add_row([
1010 0 : format!("{}", node.id),
1011 0 : node.listen_http_addr,
1012 0 : node.availability_zone_id,
1013 0 : format!("{:?}", node.scheduling),
1014 0 : format!("{:?}", node.availability),
1015 0 : ]);
1016 0 : }
1017 0 : println!("{table}");
1018 0 : }
1019 0 : Command::TenantSetTimeBasedEviction {
1020 0 : tenant_id,
1021 0 : period,
1022 0 : threshold,
1023 0 : } => {
1024 0 : vps_client
1025 0 : .set_tenant_config(&TenantConfigRequest {
1026 0 : tenant_id,
1027 0 : config: TenantConfig {
1028 0 : eviction_policy: Some(EvictionPolicy::LayerAccessThreshold(
1029 0 : EvictionPolicyLayerAccessThreshold {
1030 0 : period: period.into(),
1031 0 : threshold: threshold.into(),
1032 0 : },
1033 0 : )),
1034 0 : heatmap_period: Some(Duration::from_secs(300)),
1035 0 : ..Default::default()
1036 0 : },
1037 0 : })
1038 0 : .await?;
1039 0 : }
1040 0 : Command::BulkMigrate {
1041 0 : nodes,
1042 0 : concurrency,
1043 0 : max_shards,
1044 0 : dry_run,
1045 0 : } => {
1046 0 : // Load the list of nodes, split them up into the drained and filled sets,
1047 0 : // and validate that draining is possible.
1048 0 : let node_descs = storcon_client
1049 0 : .dispatch::<(), Vec<NodeDescribeResponse>>(
1050 0 : Method::GET,
1051 0 : "control/v1/node".to_string(),
1052 0 : None,
1053 0 : )
1054 0 : .await?;
1055 0 :
1056 0 : let mut node_to_drain_descs = Vec::new();
1057 0 : let mut node_to_fill_descs = Vec::new();
1058 0 :
1059 0 : for desc in node_descs {
1060 0 : let to_drain = nodes.contains(&desc.id);
1061 0 : if to_drain {
1062 0 : node_to_drain_descs.push(desc);
1063 0 : } else {
1064 0 : node_to_fill_descs.push(desc);
1065 0 : }
1066 0 : }
1067 0 :
1068 0 : if nodes.len() != node_to_drain_descs.len() {
1069 0 : anyhow::bail!("Bulk migration requested away from node which doesn't exist.")
1070 0 : }
1071 0 :
1072 0 : node_to_fill_descs.retain(|desc| {
1073 0 : matches!(desc.availability, NodeAvailabilityWrapper::Active)
1074 0 : && matches!(
1075 0 : desc.scheduling,
1076 0 : NodeSchedulingPolicy::Active | NodeSchedulingPolicy::Filling
1077 0 : )
1078 0 : });
1079 0 :
1080 0 : if node_to_fill_descs.is_empty() {
1081 0 : anyhow::bail!("There are no nodes to migrate to")
1082 0 : }
1083 0 :
1084 0 : // Set the node scheduling policy to draining for the nodes which
1085 0 : // we plan to drain.
1086 0 : for node_desc in node_to_drain_descs.iter() {
1087 0 : let req = NodeConfigureRequest {
1088 0 : node_id: node_desc.id,
1089 0 : availability: None,
1090 0 : scheduling: Some(NodeSchedulingPolicy::Draining),
1091 0 : };
1092 0 :
1093 0 : storcon_client
1094 0 : .dispatch::<_, ()>(
1095 0 : Method::PUT,
1096 0 : format!("control/v1/node/{}/config", node_desc.id),
1097 0 : Some(req),
1098 0 : )
1099 0 : .await?;
1100 0 : }
1101 0 :
1102 0 : // Perform the migration: move each tenant shard scheduled on a node to
1103 0 : // be drained to a node which is being filled. A simple round robin
1104 0 : // strategy is used to pick the new node.
1105 0 : let tenants = storcon_client
1106 0 : .dispatch::<(), Vec<TenantDescribeResponse>>(
1107 0 : Method::GET,
1108 0 : "control/v1/tenant".to_string(),
1109 0 : None,
1110 0 : )
1111 0 : .await?;
1112 0 :
1113 0 : let mut selected_node_idx = 0;
1114 0 :
1115 0 : struct MigrationMove {
1116 0 : tenant_shard_id: TenantShardId,
1117 0 : from: NodeId,
1118 0 : to: NodeId,
1119 0 : }
1120 0 :
1121 0 : let mut moves: Vec<MigrationMove> = Vec::new();
1122 0 :
1123 0 : let shards = tenants
1124 0 : .into_iter()
1125 0 : .flat_map(|tenant| tenant.shards.into_iter());
1126 0 : for shard in shards {
1127 0 : if let Some(max_shards) = max_shards {
1128 0 : if moves.len() >= max_shards {
1129 0 : println!(
1130 0 : "Stop planning shard moves since the requested maximum was reached"
1131 0 : );
1132 0 : break;
1133 0 : }
1134 0 : }
1135 0 :
1136 0 : let should_migrate = {
1137 0 : if let Some(attached_to) = shard.node_attached {
1138 0 : node_to_drain_descs
1139 0 : .iter()
1140 0 : .map(|desc| desc.id)
1141 0 : .any(|id| id == attached_to)
1142 0 : } else {
1143 0 : false
1144 0 : }
1145 0 : };
1146 0 :
1147 0 : if !should_migrate {
1148 0 : continue;
1149 0 : }
1150 0 :
1151 0 : moves.push(MigrationMove {
1152 0 : tenant_shard_id: shard.tenant_shard_id,
1153 0 : from: shard
1154 0 : .node_attached
1155 0 : .expect("We only migrate attached tenant shards"),
1156 0 : to: node_to_fill_descs[selected_node_idx].id,
1157 0 : });
1158 0 : selected_node_idx = (selected_node_idx + 1) % node_to_fill_descs.len();
1159 0 : }
1160 0 :
1161 0 : let total_moves = moves.len();
1162 0 :
1163 0 : if dry_run == Some(true) {
1164 0 : println!("Dryrun requested. Planned {total_moves} moves:");
1165 0 : for mv in &moves {
1166 0 : println!("{}: {} -> {}", mv.tenant_shard_id, mv.from, mv.to)
1167 0 : }
1168 0 :
1169 0 : return Ok(());
1170 0 : }
1171 0 :
1172 0 : const DEFAULT_MIGRATE_CONCURRENCY: usize = 8;
1173 0 : let mut stream = futures::stream::iter(moves)
1174 0 : .map(|mv| {
1175 0 : let client = Client::new(http_client.clone(), cli.api.clone(), cli.jwt.clone());
1176 0 : async move {
1177 0 : client
1178 0 : .dispatch::<TenantShardMigrateRequest, TenantShardMigrateResponse>(
1179 0 : Method::PUT,
1180 0 : format!("control/v1/tenant/{}/migrate", mv.tenant_shard_id),
1181 0 : Some(TenantShardMigrateRequest {
1182 0 : node_id: mv.to,
1183 0 : origin_node_id: Some(mv.from),
1184 0 : migration_config: MigrationConfig::default(),
1185 0 : }),
1186 0 : )
1187 0 : .await
1188 0 : .map_err(|e| (mv.tenant_shard_id, mv.from, mv.to, e))
1189 0 : }
1190 0 : })
1191 0 : .buffered(concurrency.unwrap_or(DEFAULT_MIGRATE_CONCURRENCY));
1192 0 :
1193 0 : let mut success = 0;
1194 0 : let mut failure = 0;
1195 0 :
1196 0 : while let Some(res) = stream.next().await {
1197 0 : match res {
1198 0 : Ok(_) => {
1199 0 : success += 1;
1200 0 : }
1201 0 : Err((tenant_shard_id, from, to, error)) => {
1202 0 : failure += 1;
1203 0 : println!(
1204 0 : "Failed to migrate {tenant_shard_id} from node {from} to node {to}: {error}"
1205 0 : );
1206 0 : }
1207 0 : }
1208 0 :
1209 0 : if (success + failure) % 20 == 0 {
1210 0 : println!(
1211 0 : "Processed {}/{} shards: {} succeeded, {} failed",
1212 0 : success + failure,
1213 0 : total_moves,
1214 0 : success,
1215 0 : failure
1216 0 : );
1217 0 : }
1218 0 : }
1219 0 :
1220 0 : println!(
1221 0 : "Processed {}/{} shards: {} succeeded, {} failed",
1222 0 : success + failure,
1223 0 : total_moves,
1224 0 : success,
1225 0 : failure
1226 0 : );
1227 0 : }
1228 0 : Command::StartDrain { node_id } => {
1229 0 : storcon_client
1230 0 : .dispatch::<(), ()>(
1231 0 : Method::PUT,
1232 0 : format!("control/v1/node/{node_id}/drain"),
1233 0 : None,
1234 0 : )
1235 0 : .await?;
1236 0 : println!("Drain started for {node_id}");
1237 0 : }
1238 0 : Command::CancelDrain { node_id, timeout } => {
1239 0 : storcon_client
1240 0 : .dispatch::<(), ()>(
1241 0 : Method::DELETE,
1242 0 : format!("control/v1/node/{node_id}/drain"),
1243 0 : None,
1244 0 : )
1245 0 : .await?;
1246 0 :
1247 0 : println!("Waiting for node {node_id} to quiesce on scheduling policy ...");
1248 0 :
1249 0 : let final_policy =
1250 0 : wait_for_scheduling_policy(storcon_client, node_id, *timeout, |sched| {
1251 0 : use NodeSchedulingPolicy::*;
1252 0 : matches!(sched, Active | PauseForRestart)
1253 0 : })
1254 0 : .await?;
1255 0 :
1256 0 : println!(
1257 0 : "Drain was cancelled for node {node_id}. Schedulling policy is now {final_policy:?}"
1258 0 : );
1259 0 : }
1260 0 : Command::StartFill { node_id } => {
1261 0 : storcon_client
1262 0 : .dispatch::<(), ()>(Method::PUT, format!("control/v1/node/{node_id}/fill"), None)
1263 0 : .await?;
1264 0 :
1265 0 : println!("Fill started for {node_id}");
1266 0 : }
1267 0 : Command::CancelFill { node_id, timeout } => {
1268 0 : storcon_client
1269 0 : .dispatch::<(), ()>(
1270 0 : Method::DELETE,
1271 0 : format!("control/v1/node/{node_id}/fill"),
1272 0 : None,
1273 0 : )
1274 0 : .await?;
1275 0 :
1276 0 : println!("Waiting for node {node_id} to quiesce on scheduling policy ...");
1277 0 :
1278 0 : let final_policy =
1279 0 : wait_for_scheduling_policy(storcon_client, node_id, *timeout, |sched| {
1280 0 : use NodeSchedulingPolicy::*;
1281 0 : matches!(sched, Active)
1282 0 : })
1283 0 : .await?;
1284 0 :
1285 0 : println!(
1286 0 : "Fill was cancelled for node {node_id}. Schedulling policy is now {final_policy:?}"
1287 0 : );
1288 0 : }
1289 0 : Command::Safekeepers {} => {
1290 0 : let mut resp = storcon_client
1291 0 : .dispatch::<(), Vec<SafekeeperDescribeResponse>>(
1292 0 : Method::GET,
1293 0 : "control/v1/safekeeper".to_string(),
1294 0 : None,
1295 0 : )
1296 0 : .await?;
1297 0 :
1298 0 : resp.sort_by(|a, b| a.id.cmp(&b.id));
1299 0 :
1300 0 : let mut table = comfy_table::Table::new();
1301 0 : table.set_header([
1302 0 : "Id",
1303 0 : "Version",
1304 0 : "Host",
1305 0 : "Port",
1306 0 : "Http Port",
1307 0 : "AZ Id",
1308 0 : "Scheduling",
1309 0 : ]);
1310 0 : for sk in resp {
1311 0 : table.add_row([
1312 0 : format!("{}", sk.id),
1313 0 : format!("{}", sk.version),
1314 0 : sk.host,
1315 0 : format!("{}", sk.port),
1316 0 : format!("{}", sk.http_port),
1317 0 : sk.availability_zone_id.clone(),
1318 0 : String::from(sk.scheduling_policy),
1319 0 : ]);
1320 0 : }
1321 0 : println!("{table}");
1322 0 : }
1323 0 : Command::SafekeeperScheduling {
1324 0 : node_id,
1325 0 : scheduling_policy,
1326 0 : } => {
1327 0 : let scheduling_policy = scheduling_policy.0;
1328 0 : storcon_client
1329 0 : .dispatch::<SafekeeperSchedulingPolicyRequest, ()>(
1330 0 : Method::POST,
1331 0 : format!("control/v1/safekeeper/{node_id}/scheduling_policy"),
1332 0 : Some(SafekeeperSchedulingPolicyRequest { scheduling_policy }),
1333 0 : )
1334 0 : .await?;
1335 0 : println!(
1336 0 : "Scheduling policy of {node_id} set to {}",
1337 0 : String::from(scheduling_policy)
1338 0 : );
1339 0 : }
1340 0 : Command::DownloadHeatmapLayers {
1341 0 : tenant_shard_id,
1342 0 : timeline_id,
1343 0 : concurrency,
1344 0 : } => {
1345 0 : let mut path = format!(
1346 0 : "v1/tenant/{tenant_shard_id}/timeline/{timeline_id}/download_heatmap_layers",
1347 0 : );
1348 0 :
1349 0 : if let Some(c) = concurrency {
1350 0 : path = format!("{path}?concurrency={c}");
1351 0 : }
1352 0 :
1353 0 : storcon_client
1354 0 : .dispatch::<(), ()>(Method::POST, path, None)
1355 0 : .await?;
1356 0 : }
1357 0 : Command::TimelineLocate {
1358 0 : tenant_id,
1359 0 : timeline_id,
1360 0 : } => {
1361 0 : let path = format!("debug/v1/tenant/{tenant_id}/timeline/{timeline_id}/locate");
1362 0 :
1363 0 : let resp = storcon_client
1364 0 : .dispatch::<(), TimelineLocateResponse>(Method::GET, path, None)
1365 0 : .await?;
1366 0 :
1367 0 : let sk_set = resp.sk_set.iter().map(|id| id.0 as i64).collect::<Vec<_>>();
1368 0 : let new_sk_set = resp
1369 0 : .new_sk_set
1370 0 : .as_ref()
1371 0 : .map(|ids| ids.iter().map(|id| id.0 as i64).collect::<Vec<_>>());
1372 0 :
1373 0 : println!("generation = {}", resp.generation);
1374 0 : println!("sk_set = {sk_set:?}");
1375 0 : println!("new_sk_set = {new_sk_set:?}");
1376 0 : }
1377 0 : Command::TimelineSafekeeperMigrate {
1378 0 : tenant_id,
1379 0 : timeline_id,
1380 0 : new_sk_set,
1381 0 : } => {
1382 0 : let path = format!("v1/tenant/{tenant_id}/timeline/{timeline_id}/safekeeper_migrate");
1383 0 :
1384 0 : storcon_client
1385 0 : .dispatch::<_, ()>(
1386 0 : Method::POST,
1387 0 : path,
1388 0 : Some(TimelineSafekeeperMigrateRequest { new_sk_set }),
1389 0 : )
1390 0 : .await?;
1391 0 : }
1392 0 : }
1393 0 :
1394 0 : Ok(())
1395 0 : }
1396 :
1397 : static WATCH_INTERVAL: Duration = Duration::from_secs(5);
1398 :
1399 0 : async fn watch_tenant_shard(
1400 0 : storcon_client: Client,
1401 0 : tenant_shard_id: TenantShardId,
1402 0 : until_migrated_to: Option<NodeId>,
1403 0 : ) -> anyhow::Result<()> {
1404 0 : if let Some(until_migrated_to) = until_migrated_to {
1405 0 : println!(
1406 0 : "Waiting for tenant shard {tenant_shard_id} to be migrated to node {until_migrated_to}"
1407 0 : );
1408 0 : }
1409 :
1410 : loop {
1411 0 : let desc = storcon_client
1412 0 : .dispatch::<(), TenantDescribeResponse>(
1413 0 : Method::GET,
1414 0 : format!("control/v1/tenant/{}", tenant_shard_id.tenant_id),
1415 0 : None,
1416 0 : )
1417 0 : .await?;
1418 :
1419 : // Output the current state of the tenant shard
1420 0 : let shard = desc
1421 0 : .shards
1422 0 : .iter()
1423 0 : .find(|s| s.tenant_shard_id == tenant_shard_id)
1424 0 : .ok_or(anyhow::anyhow!("Tenant shard not found"))?;
1425 0 : let summary = format!(
1426 0 : "attached: {} secondary: {} {}",
1427 0 : shard
1428 0 : .node_attached
1429 0 : .map(|n| format!("{n}"))
1430 0 : .unwrap_or("none".to_string()),
1431 0 : shard
1432 0 : .node_secondary
1433 0 : .iter()
1434 0 : .map(|n| n.to_string())
1435 0 : .collect::<Vec<_>>()
1436 0 : .join(","),
1437 0 : if shard.is_reconciling {
1438 0 : "(reconciler active)"
1439 : } else {
1440 0 : "(reconciler idle)"
1441 : }
1442 : );
1443 0 : println!("{summary}");
1444 :
1445 : // Maybe drop out if we finished migration
1446 0 : if let Some(until_migrated_to) = until_migrated_to {
1447 0 : if shard.node_attached == Some(until_migrated_to) && !shard.is_reconciling {
1448 0 : println!("Tenant shard {tenant_shard_id} is now on node {until_migrated_to}");
1449 0 : break;
1450 0 : }
1451 0 : }
1452 :
1453 0 : tokio::time::sleep(WATCH_INTERVAL).await;
1454 : }
1455 0 : Ok(())
1456 0 : }
|
