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