[UA] Tight worker loop can cause high CPU usage (#60950)

* Addded worker padding to save some CPU

* Updated comments

* Update worker scheduler and add a new util

The worker scheduler should only sleep when it cannot process any
in progress operations. Additionally, logic has been added
for handling of queue operations that have been in the queue for
a long time and may be viewed as still in small window of time
by wokers that do not have the credentials to process those
reindex operations.

* res 👉🏻resolve

Co-authored-by: Elastic Machine <elasticmachine@users.noreply.github.com>
This commit is contained in:
Jean-Louis Leysens 2020-03-31 17:26:46 +02:00 committed by GitHub
parent 069439a7ec
commit 12c8ff78d7
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
2 changed files with 37 additions and 7 deletions

View file

@ -50,6 +50,9 @@ const orderQueuedReindexOperations = ({
),
});
export const isQueuedOp = (op: ReindexSavedObject) =>
Boolean(op.attributes.reindexOptions?.queueSettings);
export const queuedOpHasStarted = (op: ReindexSavedObject) =>
Boolean(op.attributes.reindexOptions?.queueSettings?.startedAt);

View file

@ -10,12 +10,18 @@ import { Credential, CredentialStore } from './credential_store';
import { reindexActionsFactory } from './reindex_actions';
import { ReindexService, reindexServiceFactory } from './reindex_service';
import { LicensingPluginSetup } from '../../../../licensing/server';
import { sortAndOrderReindexOperations, queuedOpHasStarted } from './op_utils';
import { sortAndOrderReindexOperations, queuedOpHasStarted, isQueuedOp } from './op_utils';
const POLL_INTERVAL = 30000;
// If no nodes have been able to update this index in 2 minutes (due to missing credentials), set to paused.
const PAUSE_WINDOW = POLL_INTERVAL * 4;
/**
* To avoid running the worker loop very tightly and causing a CPU bottleneck we use this
* padding to simulate an asynchronous sleep. See the description of the tight loop below.
*/
const WORKER_PADDING_MS = 1000;
/**
* A singleton worker that will coordinate two polling loops:
* (1) A longer loop that polls for reindex operations that are in progress. If any are found, loop (2) is started.
@ -102,16 +108,25 @@ export class ReindexWorker {
/**
* Runs an async loop until all inProgress jobs are complete or failed.
*/
private startUpdateOperationLoop = async () => {
private startUpdateOperationLoop = async (): Promise<void> => {
this.updateOperationLoopRunning = true;
try {
while (this.inProgressOps.length > 0) {
this.log.debug(`Updating ${this.inProgressOps.length} reindex operations`);
// Push each operation through the state machine and refresh.
await Promise.all(this.inProgressOps.map(this.processNextStep));
await this.refresh();
if (
this.inProgressOps.length &&
this.inProgressOps.every(op => !this.credentialStore.get(op))
) {
// TODO: This tight loop needs something to relax potentially high CPU demands so this padding is added.
// This scheduler should be revisited in future.
await new Promise(resolve => setTimeout(resolve, WORKER_PADDING_MS));
}
}
} finally {
this.updateOperationLoopRunning = false;
@ -173,20 +188,32 @@ export class ReindexWorker {
}
};
private processNextStep = async (reindexOp: ReindexSavedObject) => {
private lastCheckedQueuedOpId: string | undefined;
private processNextStep = async (reindexOp: ReindexSavedObject): Promise<void> => {
const credential = this.credentialStore.get(reindexOp);
if (!credential) {
// Set to paused state if the job hasn't been updated in PAUSE_WINDOW.
// If this is a queued reindex op, and we know there can only ever be one in progress at a
// given time, there is a small chance it may have just reached the front of the queue so
// we give it a chance to be updated by another worker with credentials by making this a
// noop once. If it has not been updated by the next loop we will mark it paused if it
// falls outside of PAUSE_WINDOW.
if (isQueuedOp(reindexOp)) {
if (this.lastCheckedQueuedOpId !== reindexOp.id) {
this.lastCheckedQueuedOpId = reindexOp.id;
return;
}
}
// This indicates that no Kibana nodes currently have credentials to update this job.
const now = moment();
const updatedAt = moment(reindexOp.updated_at);
if (updatedAt < now.subtract(PAUSE_WINDOW)) {
return this.reindexService.pauseReindexOperation(reindexOp.attributes.indexName);
await this.reindexService.pauseReindexOperation(reindexOp.attributes.indexName);
return;
} else {
// If it has been updated recently, we assume another node has the necessary credentials,
// and this becomes a noop.
return reindexOp;
return;
}
}