939255 - Releng db servers in scl3

Status: RESOLVED FIXED

(Infrastructure & Operations Graveyard :: WebOps: Other, task)

Description

[Sheeri Cabral [:sheeri]]

original purpose of releng db machines:

We need some new databases for the new releng cluster in scl3. These will be used for     tbpl, ship-it, sheriffs, treestatus, etc which will move from genericrhel6 in phx1 to relengweb in scl3.

############################################################
The machines were built in bug 861369. and then I asked:
:jd - Should these be master/master or master/slave? Can the "ship-it, sheriffs, treestatus, etc" applications handle autoincrements that increment by 2 instead of 1? If not, they will be master/slave.

Also, will tbpl still be moving over, or is it happy where it is on its own server in phx?
############################################################

I have no idea how the applications are coded. Actually I forgot about this project and heard recently that sheriffs was dead or dying.

I think these are basic web sites, but cant speak to anything more than that.

I think the plan is still to move tbpl. The original idea was to get all the tree closing stuff in one place (or at least off the generic services) to reduce tree closing events.

Jake may know more but I imagine the devs would need to be consulted regarding the ability of the apps to increment by 2.

############################################################

So this bug is to deal with these questions.

Comment 1

[Sheeri Cabral [:sheeri]]

Putting the needinfo in here to see if Jake has any clue about the above; also if he knows who the devs to contact would be.

Comment 2

[Amy Rich [:arr] [:arich]]

Tagging coop with a needinfo since he may have context on the last question about who the "devs" are to talk to about at least some of these databases.

Comment 3

[Chris Cooper [:coop] (he/him)]

None of the releng apps make assumptions about how auto-incrementing happens. We're good here.

Comment 4

[Sheeri Cabral [:sheeri]]

OK, so looks like we can make it a master/master database then.

Comment 5

[Jason Crowe [:jd]]

looks like this bug is no longer necessary as the dependent bug is r/f.

Comment 6

[Sheeri Cabral [:sheeri]]

I don't see a dependent bug here, reopening. 

Do we need to plan on when the move is, or is this something that releng has a timeline for?

Comment 7

[Jason Crowe [:jd]]

okay, it was my understanding that this bug was to answer the questions raised in 861369 which I called dependent, but if you wish to be pedantic it is not in that field.

That bug was to resolve one of the issues prior to moving TBPL to the releng cluster in scl3 in bug 860381. That bug was closed when a temporary solution was in place.

I guess this bug can be morphed into something but at the moment I do not recognize any actionable items listed herein as the questions stated here have been answered.

Comment 8

[Jason Crowe [:jd]]

oh, also I am not aware of any timeline. I was under the impression that webops would get suckered into migrating this stuff anyhow. Jake would probably be the one with those answers.

Comment 9

[Sheeri Cabral [:sheeri]]

My intention is "it's made, please use it and don't forget that we built it for you".

Comment 10

[Sheeri Cabral [:sheeri]]

Any objection to virtualizing these servers, since they're not actually being used yet?

Comment 11

[Jason Crowe [:jd]]

Nope, go for it :)

Thanks

Comment 12

[Amy Rich [:arr] [:arich]]

Is there a project in progress to move these?  If not, should we just decom them until there is an active project to do this work?

Comment 13

[Jason Crowe [:jd]]

(In reply to Amy Rich [:arich] [:arr] from comment #12)
> Is there a project in progress to move these?  If not, should we just decom
> them until there is an active project to do this work?

yes, the biggest thing is tbpl which will go to the releng web cluster there and needs these databases.

Bug is here 983667

as far as when this work will be scheduled I am not sure. We (webops) are working on better scheduling and planning, but it is a process.

Comment 14

[Sheeri Cabral [:sheeri]]

We have decided that we can live with out-of-warranty hardware on tbpl until treeherder takes over towards the end of the year (scheduled for q3).

They were spun up in bug 998491. I will update bug 983667.

Sounds like decom'ing the releng servers is the way to go; thanx!

Comment 15

[Jason Crowe [:jd]]

Actualy no. THis is about moving all the tree closing stuff out of phx1 as most of it is in scl3 now.

Comment 16

[Sheeri Cabral [:sheeri]]

Hrm, we can't virtualize these if they're going to house tbpl db's, those dbs are too resource intensive. I will defer to Sean to see if we can work something out.

Comment 17

[Jason Crowe [:jd]]

Okay, we discussed it and are fine with tbpl staying in phx1 till EOL. These servers can be decommissioned now.

Thanks

Comment 18

[Sheeri Cabral [:sheeri]]

Thanx! glad it's all worked out, and hopefully not moving before EOL means less work all around for everyone.