Talking to a couple elasticsearch engineers at elasticon one suggestion for how to deal with our master node timeout on cluster state update issues was to compact multiple small indices into a single index. The suggested method was to use different field names for each type, such as title_aawiki, title_abwiki, etc. etc. This would allow to massively shrink the number of indices and shards the elasticsearch master needs to keep track of.
Description
Description
Event Timeline
Restricted Application added a project: Discovery-ARCHIVED. · View Herald TranscriptFeb 18 2016, 12:53 AM2016-02-18 00:53:05 (UTC+0)
• Deskana moved this task from Inbox to Technical on the CirrusSearch board.Feb 18 2016, 4:10 AM2016-02-18 04:10:29 (UTC+0)
MaxSlo moved this task from Technical to Multilingual and cross-project on the CirrusSearch board.Feb 23 2021, 4:44 PM2021-02-23 16:44:09 (UTC+0)
Restricted Application added a project: Discovery-Search. · View Herald TranscriptFeb 23 2021, 4:44 PM2021-02-23 16:44:14 (UTC+0)
MaxSlo moved this task from Multilingual and cross-project to Technical on the CirrusSearch board.Feb 23 2021, 4:44 PM2021-02-23 16:44:21 (UTC+0)
Gehel subscribed.
Comment ActionsThe approach we've taken is to create multiple clusters, to this task does not make sense anymore.