Page MenuHomePhabricator

SUP: Shift Writes from Cirrus to SUP
Closed, ResolvedPublic

Description

As a final step, we want to shift writes from Cirrus to SUP. To reduce the risk of loosing data, we take the following measures:

  • only shift writes for public wikis
  • only shift writes for one DC at a time
  • only shift writes for a subset of wikis at a time

The batches are:

  1. testwiki; frwiki; itwiki; commonswiki; wikidatawiki (DC: CODFW)
  2. non-private rest (DC: CODFW)

AC:

  1. wmf-config: wgCirrusSearchWriteClusters reflects those changes
  2. SUP release consumer-search: --wikiids reflects those changes

Event Timeline

Change #1024411 had a related patch set uploaded (by Peter Fischer; author: Peter Fischer):

[operations/mediawiki-config@master] Shift writes to SUP, 1st batch

https://gerrit.wikimedia.org/r/1024411

Change #1025411 had a related patch set uploaded (by Ebernhardson; author: Ebernhardson):

[operations/deployment-charts@master] cirrus updater: Enable codfw consumer

https://gerrit.wikimedia.org/r/1025411

Change #1025411 merged by jenkins-bot:

[operations/deployment-charts@master] cirrus updater: Enable codfw consumer

https://gerrit.wikimedia.org/r/1025411

Change #1024411 merged by jenkins-bot:

[operations/mediawiki-config@master] Shift writes to SUP, 1st batch

https://gerrit.wikimedia.org/r/1024411

Mentioned in SAL (#wikimedia-operations) [2024-04-29T20:05:39Z] <cjming@deploy1002> Started scap: Backport for [[gerrit:1024411|Shift writes to SUP, 1st batch (T363475)]]

Mentioned in SAL (#wikimedia-operations) [2024-04-29T20:09:10Z] <cjming@deploy1002> cjming and pfischer: Backport for [[gerrit:1024411|Shift writes to SUP, 1st batch (T363475)]] synced to the testservers (https://wikitech.wikimedia.org/wiki/Mwdebug)

Mentioned in SAL (#wikimedia-operations) [2024-04-29T20:21:56Z] <cjming@deploy1002> Finished scap: Backport for [[gerrit:1024411|Shift writes to SUP, 1st batch (T363475)]] (duration: 16m 17s)

Change #1029273 had a related patch set uploaded (by Ebernhardson; author: Ebernhardson):

[operations/mediawiki-config@master] cirrus: Shift remaining public wikis in codfw to replacement updater

https://gerrit.wikimedia.org/r/1029273

Change #1029274 had a related patch set uploaded (by Ebernhardson; author: Ebernhardson):

[operations/deployment-charts@master] cirrus: Expand codfw to serve writes to all public wikis

https://gerrit.wikimedia.org/r/1029274

Change #1029274 merged by jenkins-bot:

[operations/deployment-charts@master] cirrus: Expand codfw to serve writes to all public wikis

https://gerrit.wikimedia.org/r/1029274

Change #1029273 merged by jenkins-bot:

[operations/mediawiki-config@master] cirrus: Shift remaining public wikis in codfw to replacement updater

https://gerrit.wikimedia.org/r/1029273

Mentioned in SAL (#wikimedia-operations) [2024-05-08T20:09:54Z] <ebernhardson@deploy1002> Started scap: Backport for [[gerrit:1029273|cirrus: Shift remaining public wikis in codfw to replacement updater (T363475)]]

Mentioned in SAL (#wikimedia-operations) [2024-05-08T20:12:34Z] <ebernhardson@deploy1002> ebernhardson: Backport for [[gerrit:1029273|cirrus: Shift remaining public wikis in codfw to replacement updater (T363475)]] synced to the testservers (https://wikitech.wikimedia.org/wiki/Mwdebug)

Mentioned in SAL (#wikimedia-operations) [2024-05-08T20:25:54Z] <ebernhardson@deploy1002> Finished scap: Backport for [[gerrit:1029273|cirrus: Shift remaining public wikis in codfw to replacement updater (T363475)]] (duration: 16m 00s)

pfischer moved this task from Incoming to In Progress on the Discovery-Search (Current work) board.
pfischer updated Other Assignee, added: EBernhardson.
dr0ptp4kt subscribed.

Peter's on a WDQS task, Erik will take this after the Discolytics search metrics task.

Change #1031027 had a related patch set uploaded (by Ebernhardson; author: Ebernhardson):

[operations/deployment-charts@master] cirrus: Deploy updater to eqiad at 25% load

https://gerrit.wikimedia.org/r/1031027

Change #1031029 had a related patch set uploaded (by Ebernhardson; author: Ebernhardson):

[operations/mediawiki-config@master] cirrus: Shift 25% of public wikis writes in eqiad to replacement updater

https://gerrit.wikimedia.org/r/1031029

Change #1031027 merged by jenkins-bot:

[operations/deployment-charts@master] cirrus: Deploy updater to eqiad at 25% load

https://gerrit.wikimedia.org/r/1031027

Change #1031029 merged by jenkins-bot:

[operations/mediawiki-config@master] cirrus: Shift 25% of public wikis writes in eqiad to replacement updater

https://gerrit.wikimedia.org/r/1031029

Mentioned in SAL (#wikimedia-operations) [2024-05-14T20:26:22Z] <cjming@deploy1002> Started scap: Backport for [[gerrit:1031029|cirrus: Shift 25% of public wikis writes in eqiad to replacement updater (T363475)]]

Mentioned in SAL (#wikimedia-operations) [2024-05-14T20:28:58Z] <cjming@deploy1002> cjming and ebernhardson: Backport for [[gerrit:1031029|cirrus: Shift 25% of public wikis writes in eqiad to replacement updater (T363475)]] synced to the testservers (https://wikitech.wikimedia.org/wiki/Mwdebug)

Mentioned in SAL (#wikimedia-operations) [2024-05-14T20:41:25Z] <cjming@deploy1002> Finished scap: Backport for [[gerrit:1031029|cirrus: Shift 25% of public wikis writes in eqiad to replacement updater (T363475)]] (duration: 15m 02s)

Change #1032017 had a related patch set uploaded (by Ebernhardson; author: Ebernhardson):

[mediawiki/extensions/CirrusSearch@master] Switch DefaultWriteClusters settings to provide_default

https://gerrit.wikimedia.org/r/1032017

Change #1032017 abandoned by Ebernhardson:

[mediawiki/extensions/CirrusSearch@master] Switch DefaultWriteClusters settings to provide_default

Reason:

On review of how extension registration works, this is simply not possible. With the way values are passed around, it looks like provide_default can only provide arrays. This is being fixed via
Ia3eaf6a6078 instead.

https://gerrit.wikimedia.org/r/1032017

We are getting alerts from "MediaWiki CirrusSearch update rate - codfw". I've silenced them for the moment, but we might want to make sure we have the appropriate alerts for the new SUP and that the old one does not raise false positives.

Change #1036723 had a related patch set uploaded (by Ebernhardson; author: Ebernhardson):

[operations/mediawiki-config@master] cirrus: Move remaining public writes to SUP

https://gerrit.wikimedia.org/r/1036723

Change #1036723 merged by jenkins-bot:

[operations/mediawiki-config@master] cirrus: Move remaining public writes to SUP

https://gerrit.wikimedia.org/r/1036723

Mentioned in SAL (#wikimedia-operations) [2024-05-28T20:22:20Z] <cjming@deploy1002> Started scap: Backport for [[gerrit:1036723|cirrus: Move remaining public writes to SUP (T363475)]]

Mentioned in SAL (#wikimedia-operations) [2024-05-28T20:25:26Z] <cjming@deploy1002> cjming and ebernhardson: Backport for [[gerrit:1036723|cirrus: Move remaining public writes to SUP (T363475)]] synced to the testservers (https://wikitech.wikimedia.org/wiki/Mwdebug)

Mentioned in SAL (#wikimedia-operations) [2024-05-28T20:34:31Z] <cjming@deploy1002> Finished scap: Backport for [[gerrit:1036723|cirrus: Move remaining public writes to SUP (T363475)]] (duration: 12m 11s)