In order to create a mediawiki stack in codfw, we need to set up the redis clusters there. We actually have two kind-of separate clusters in eqiad, rdb* and rbf*
Description
Details
Subject | Repo | Branch | Lines +/- | |
---|---|---|---|---|
fix rdb codfw hostnames in netboot.cfg | operations/puppet | production | +1 -1 |
Status | Subtype | Assigned | Task | ||
---|---|---|---|---|---|
Resolved | Joe | T86894 Set up the mediawiki application layer in codfw | |||
Resolved | Joe | T86887 Setup redis clusters in codfw | |||
Resolved | Dzahn | T86898 Check that the redis roles can be applied in codfw, set up puppet. | |||
Resolved | Dzahn | T86897 Procure and setup rbf2001-2002 | |||
Resolved | Papaul | T86940 label & setup drac/basic setings for rbf2001 & rbf2002 | |||
Resolved | Papaul | T88380 reclaim rbf2002/WMF5833 back to spare, allocate WMF5845 as rbf2002 | |||
Resolved | RobH | T86896 Procure rdb2001-2004 - onsite pending racking |
Event Timeline
Change 188227 had a related patch set uploaded (by Dzahn):
add rbf2001/2002 to site.pp
@Joe how about master-slave replication of redis servers? i see the "# slaveof <masterip> <masterport>" construct in the config template. Is the goal to just replicate within a datacenter? so this should be an exact copy of eqiad but we never want to replicate cross datacenters?
At the moment we don't plan to replicate between datacenters. Maybe just the sessions data, in case, but multi-dc replication with redis is not an easy task for sure.
We are going with jessie.
< _joe_> so yes, let's go with jessie
< _joe_> we sould also look at changes in redis between the version we have and the one on jessie
Change 196104 had a related patch set uploaded (by Dzahn):
fix rdb codfw hostnames in netboot.cfg