Closed
Bug 813317
Opened 12 years ago
Closed 12 years ago
Reporting replica for Socorro
Categories
(Data & BI Services Team :: DB: MySQL, task)
Tracking
(Not tracked)
RESOLVED
FIXED
People
(Reporter: selenamarie, Unassigned)
References
Details
We've got a replica for Socorro's primary database, and have been directing some users to run read-only queries against it.
If this service becomes more popular, the chance that a long-running query causes the replica to go out of sync with the master database increases. Replication in Postgres can be delayed by certain read-only queries because of lock contention.
To avoid a production problem, it seems wise to have an additional replica dedicated to running queries against, that is configured to recover from prolonged replication delay and is not considered a primary failover target.
Comment 1•12 years ago
|
||
What are the desired specs for the box? disk size, SSD vs. spinning disks, RAM size, CPU speed / # CPUs?
Comment 2•12 years ago
|
||
2.5 Tb spinning disk after RAIDing, whatever CPUs come with it (quad or 8-core is plenty), at least 12 Gb RAM but 24 or higher would be nice. This is an 800Gb database, so more RAM=better.
Comment 3•12 years ago
|
||
(In reply to Sheeri Cabral [:sheeri] from comment #2)
> 2.5 Tb spinning disk after RAIDing, whatever CPUs come with it (quad or
> 8-core is plenty), at least 12 Gb RAM but 24 or higher would be nice. This
> is an 800Gb database, so more RAM=better.
The max you can get for blades is 1TB disk. Is that sufficient?
Comment 4•12 years ago
|
||
Well, the db itself is 800 Gb right now, so I don't think limiting ourselves to 1Tb is a good idea.
What's the alternatives...seamicros? Have we ever tested postgres on seamicros (we haven't in the year or so I've been here, to my knowledge).
Updated•12 years ago
|
Whiteboard: [waiting on hardware]
Updated•12 years ago
|
Whiteboard: [waiting on hardware] → [2013q1][waiting on hardware]
Comment 5•12 years ago
|
||
Corey will order an HP system for this.
Comment 6•12 years ago
|
||
We ordered 2 "sandbox" pg nodes for socorro, we can use one of those for this (should be installed this week) and I am ordering a third to backfill that one.
Comment 7•12 years ago
|
||
Bacfill shipped to Phoenix. here is tracking
Storage Blade Fed Ex 017136230767311 ETA 1/25
Hard drives (need to be installed) shipped via On Trac ETA 1/22
Server Blade shipped via Fed Ex 013840237703237 ETA 1/24
Comment 8•12 years ago
|
||
The blade server and storage blade has been installed here:
https://inventory.mozilla.org/en-US/systems/racks/?location=12&status=&rack=280&allocation=
Asset #'s:
15536
15535
Comment 9•12 years ago
|
||
closing this out as we have them being kicked in another bug. Thanks all!
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → FIXED
Updated•12 years ago
|
Whiteboard: [2013q1][waiting on hardware]
Updated•10 years ago
|
Product: mozilla.org → Data & BI Services Team
You need to log in
before you can comment on or make changes to this bug.
Description
•