The archive type has to be removed prior to upgrading to 6.x.
As we are moving to have multiple elastic instance per node we can create many more indices so the easiest solution is to simply create a new index dedicate to archives.
Description
Details
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | EBernhardson | T183281 [epic] ELK upgrade to 6.x (elasticsearch, kibana, logstash) | |||
Resolved | None | T183282 [epic] Search cluster upgrade to 6.x | |||
Resolved | None | T194199 [Epic] Prepare for Elasticsearch 6 upgrade | |||
Resolved | dcausse | T200198 Drop the "archive" type in the general index | |||
Resolved | EBernhardson | T213851 Drop "archive" from generic index on testwiki |
Event Timeline
Change 447641 had a related patch set uploaded (by DCausse; owner: DCausse):
[mediawiki/extensions/CirrusSearch@master] Add the "archive" alias
Change 447642 had a related patch set uploaded (by DCausse; owner: DCausse):
[mediawiki/extensions/CirrusSearch@master] Move "archive" to its own index
Change 448000 had a related patch set uploaded (by DCausse; owner: DCausse):
[mediawiki/extensions/CirrusSearch@master] Start using the archive alias
Change 447641 merged by jenkins-bot:
[mediawiki/extensions/CirrusSearch@master] Add the "archive" alias
Moving to waiting/blocked, I'm thinking we want to delay creating the archive indices until after we deploy multiple elasticsearch instances? We probably could deploy it before hand, the master node latency evaluation was reasonable, not great but reasonable, with only the new archive indices added. With 2x the archive indices added was where we started seeing more erratic master latency behaviour.
Safest though would be to let this be waiting/blocked until then.
Change 448000 merged by jenkins-bot:
[mediawiki/extensions/CirrusSearch@master] Start using the archive alias
Change 447642 merged by jenkins-bot:
[mediawiki/extensions/CirrusSearch@master] Move "archive" to its own index