Short downtime for s2 today - Toolserver-l - lists.wikimedia.org

List overview All Threads
Download

Short downtime for s2 today

enwiki_p

Restart of mysql on cassia (s5)

DaB.

14 Feb 2013 14 Feb '13

1:03 p.m.

Hello all, to update a package with a security-problem, I have to restart z-dat-s2-b today. This will cause sql-s2-user and sql-s2-rr to be down. The downtime wll start at TODAY, 19:20 UTC and should not more than a few minutes. Sincerely, DaB. -- Userpage: [[:w:de:User:DaB.]] — PGP: 0x2d3ee2d42b255885

Attachments:

signature.asc (application/pgp-signature — 198 bytes)

Reply

Show replies by date

DaB.

14 Feb 14 Feb

1:23 p.m.

Forgot the ticket-ID: MNT-1291 [1]. [1] https://jira.toolserver.org/browse/MNT-1291 -- Userpage: [[:w:de:User:DaB.]] — PGP: 0x2d3ee2d42b255885

Reply

DaB.

7:30 p.m.

Hello, At Thursday 14 February 2013 20:30:28 DaB. wrote:

TODAY, 19:20 UTC and should not more than a few minutes.

ok, its done. Sincerely, DaB. -- Userpage: [[:w:de:User:DaB.]] — PGP: 0x2d3ee2d42b255885

Reply

Maciej Jaros

15 Feb 15 Feb

6:38 p.m.

DaB. (2013-02-14 20:30):

Hello, At Thursday 14 February 2013 20:30:28 DaB. wrote:

TODAY, 19:20 UTC and should not more than a few minutes.

ok, its done.

Seems like S2 is going crazy for few days: http://toolserver.org/~bryan/stats/replag/?cluster=s2#s2-daily Not sure if related, but at least the restart seem not to have helped ;-). I've even received the Query-Killer greetings for the first time and what was more interesting, it was on a query that should take... hm... 0 s as it should only work on indexes (rev_pag, rev_id). Regards, Nux / eccenux.

Reply

DaB.

8:38 p.m.

Hello, At Friday 15 February 2013 21:32:53 DaB. wrote:

Seems like S2 is going crazy for few days: http://toolserver.org/~bryan/stats/replag/?cluster=s2#s2-daily Not sure if related, but at least the restart seem not to have helped ;-). I've even received the Query-Killer greetings for the first time and what was more interesting, it was on a query that should take... hm... 0 s as it should only work on indexes (rev_pag, rev_id).

I admit that there is still room for optimization (z-dat-s2-b is our first Debian-db-server and I started with a clean config), but if you look at the monthly or yearly graph you can see that there were also these peaks before. I will look what I can do, but there are several things I have to care for in parallel at the moment. Another news: I got a fresh s5-dump and will setup another s5-instance in the next days. Expect a read-only-time of s5 next week (separate mail will follow). Sincerely, DaB. -- Userpage: [[:w:de:User:DaB.]] — PGP: 0x2d3ee2d42b255885

Reply

Platonides

8:51 p.m.

On 15/02/13 21:38, DaB. wrote:

I admit that there is still room for optimization (z-dat-s2-b is our first Debian-db-server and I started with a clean config), but if you look at the monthly or yearly graph you can see that there were also these peaks before. I will look what I can do, but there are several things I have to care for in parallel at the moment.

An option would be to clone the configuration used for wmf servers. It must be puppetized, and it should apply quite well into debian.

Reply

Maciej Jaros

22 Feb 22 Feb

2:52 p.m.

DaB. (2013-02-15 21:38):

Hello, At Friday 15 February 2013 21:32:53 DaB. wrote:

Seems like S2 is going crazy for few days: http://toolserver.org/~bryan/stats/replag/?cluster=s2#s2-daily Not sure if related, but at least the restart seem not to have helped ;-). I've even received the Query-Killer greetings for the first time and what was more interesting, it was on a query that should take... hm... 0 s as it should only work on indexes (rev_pag, rev_id).

I admit that there is still room for optimization (z-dat-s2-b is our first Debian-db-server and I started with a clean config), but if you look at the monthly or yearly graph you can see that there were also these peaks before. I will look what I can do, but there are several things I have to care for in parallel at the moment.

Hmm... I admit that there were peaks before, but this are not peaks[1] it's a whole mountain range ;-). Previously there were peaks, some rumbling and it went quiet. Currently it's just a crazy ride and even queries like this can take seconds to complete (I know because this the only query on my first page): SELECT min(rc_timestamp) as rc_min, max(rc_timestamp) as rc_max FROM `recentchanges` Something must have changed. Some tool started going crazy or some device is faulty. [1] = <http://toolserver.org/~bryan/stats/replag/?cluster=s2#s2-monthly>

Reply

4079

days inactive

4087

days old

toolserver-l@lists.wikimedia.org

Manage subscription

6 comments

3 participants

tags (0)

participants (3)

DaB.
Maciej Jaros
Platonides