Elasticsearch is running happily on all the wikis. It's time to say goodbye to lsearchd.
Use "kill-lsearchd" as your topic in Gerrit.
Elasticsearch is running happily on all the wikis. It's time to say goodbye to lsearchd.
Use "kill-lsearchd" as your topic in Gerrit.
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | RobH | T85009 Decommission lsearchd | |||
Resolved | RobH | T86149 reclaim lsearchd hosts | |||
Resolved | • Cmjohnson | T92434 wipe search* and searchidx* hosts | |||
Resolved | fgiunchedi | T86150 remove lsearchd support from puppet | |||
Resolved | • Cmjohnson | T88352 wipe lsearch machines |
Can we also re-commission servers? we do have a few applications where this could be useful.
next steps:
I was under the impression that we are deprecating all pools and thus all of lsearchd?
what's a likely eta for decommissioning all the pools? if it isn't that far out I'd prefer to do this piecemeal
all backend hosts commented in pybal config for search_pool1-5 and search_prefix and disabled notification in icinga
Change 183462 had a related patch set uploaded (by Filippo Giunchedi):
lsearchd: remove lvs configuration
Change 183469 had a related patch set uploaded (by Filippo Giunchedi):
lsearchd: remove udp2log configuration
Change 184611 had a related patch set uploaded (by Filippo Giunchedi):
lsearchd: remove all lvs references
Change 184624 had a related patch set uploaded (by Filippo Giunchedi):
remove service endpoints for lsearchd
see https://gerrit.wikimedia.org/r/#/c/184620/ for puppet decom and https://gerrit.wikimedia.org/r/#/c/184624/ for dns
pending machine deprovisioning, @mark did we have an use for those already or back to the spare pool?
Depends on the age of the machines as well. IIRC some were older/out of warranty, some fairly recent. Has anyone looked at that yet?
according to the dell website warranty expired on 2014-02-02 for all machines, I couldn't find any machine with a longer warranty
curiously enough I couldn't find the asset tag for search1014 via dmidecode, racktables had it though
Then we shouldn't repurpose it for any critical roles. If anyone has any purpose for them where machine failures/lack of support isn't a problem we can keep them, but keep in mind that these won't live much longer.
understood re: machines, I can't think of a use case right now where we'd be happy with out of warranty machines. specs are 48G ram, 2x12 cores and 300GB SSD, I'll let @RobH comment on putting them back in the spare pool/shutdown/reuse etc
I'll keep this ticket assgined to me, but put in sub-tasks for the clearing of data and such.
I'm not sure if we want to keep these at all, but they certainly need the disks wiped.
I think we want to keep them as spares unless we have a good reason not to (e.g. lack of rack space, cleanup, etc) since they are fairly powerful but out of warranty
with adding the old systems back to spares, I think that ends the decom steps for lsearch. Resolving.