Minutes :: Weekly infra meeting :: 2013-01-09

Karsten 'quaid' Wade kwade at redhat.com
Thu Jan 10 07:30:59 UTC 2013


The bot wasn't yet working on Monday morning, instead I'm including the
entire logs below. I've also extracted the key points above here where
the bot-style commands were used for that purpose.

Topic:
* New hosting plan draft:
  http://etherpad.ovirt.org/p/new_hosting_design_Jan_2013
** INFO
*** Pulled together from previous and existing discussions
** ACTIONS
*** Team needs to start writing up initial Puppet manifests
* Trac review:
  https://fedorahosted.org/ovirt/report/1
** INFO
*** No time to review tickets in meeting
** ACTIONS
*** Everyone update your own tickets


## begin IRC log
07:00 < quaid> #startmeeting
07:01 < quaid> ruh roh
07:01 -!- Irssi: #ovirt: Total of 156 nicks [2 ops, 0 halfops, 0 voices,
154 normal]
07:01  * quaid goes to look at bot one moment
07:01 -!- ofri_ [~ofri at nat-pool-tlv-u1.redhat.com] has joined #ovirt
07:01 -!- ykaul [~ykaul at nat-pool-tlv-u1.redhat.com] has joined #ovirt
07:01 -!- mskrivanek|away is now known as mskrivanek
07:02 < quaid> ok, bot is down
07:02 < quaid> I was having troubles before the holidays, still not
fixed I reckon
07:02 -!- dcaro [~dcaro at nat-pool-brq-u.redhat.com] has joined #ovirt
07:02 < quaid> so I'll just keep a copy of the log here and we'll just
use meetbot convention to make searching the logs easier :/
07:03 < quaid> #meetingname Infra weekly meeting
07:03 < quaid> #topic roll call & coffee time
07:03 -!- eedri_ [~eedri at nat-pool-tlv-t1.redhat.com] has quit [Read
error: Operation timed out]
07:04 -!- gchaplik [~gchaplik at nat-pool-tlv-t1.redhat.com] has quit [Read
error: Operation timed out]
07:04  * quaid is here & pulling up agenda
07:04  * dneary is here
07:04 < dneary> Any news on the git issue earlier? /me asks quaid to add
it to agenda
07:05 < quaid> no news that I've seen, we can add it
07:07 < quaid> ok, I've read the news & am not having a problem with
gerrit access from here either
07:07 -!- mdday [~mdday at cpe-076-182-073-192.nc.res.rr.com] has joined #ovirt
07:07 < quaid> Agenda:
07:08 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has joined #ovirt
07:08 -!- danken [~danken at bzq-109-67-31-252.red.bezeqint.net] has joined
#ovirt
07:08 < quaid> Review & discuss & improve upon draft:
http://etherpad.ovirt.org/p/new_hosting_design_Jan_2013
07:08 < quaid> Trac ticket review: https://fedorahosted.org/ovirt/report/1
07:08 < quaid> Is gerrit.ovirt.org OK? Status
07:08 < quaid> All other business
07:08 < quaid> anything else?
07:09 -!- ykaul [~ykaul at nat-pool-tlv-u1.redhat.com] has quit [Ping
timeout: 480 seconds]
07:09 -!- ircuser-1 [~ircuser-1 at 35.222-62-69.ftth.swbr.surewest.net] has
joined #ovirt
07:09 -!- jclift [~jclift at nat-pool-brq-t.redhat.com] has joined #ovirt
07:10 < jclift> Is anyone around who can guide me in how to successfully
submit patches with gerrit?
07:11 < jclift> I've been following the instructions on the "Working
with oVirt Gerrit" wiki page, but it's not working properly.  Only the
first commit shows up in the web interface.
07:12 -!- abaron [~abaron at 87.68.20.33.cable.012.net.il] has quit [Ping
timeout: 480 seconds]
07:12 < quaid> ok, I'm giving up on the bot for the moment, sorry :)
07:12  * quaid getting coffee, too
07:12 < ewoud> jclift: how did you push? git push or git-review?
07:12 < quaid> #topic New hosting plan draft:
http://etherpad.ovirt.org/p/new_hosting_design_Jan_2013
07:13 < quaid> #info Pulled together from previous and existing discussions
07:13 < jclift> ewoud: git push gerrit.ovirt.org:ovirt-engine
HEAD:refs/for/master
07:14 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has quit [Quit:
Leaving.]
07:15 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has joined #ovirt
07:15 -!- aglitke [~aglitke at 68-115-116-10.dhcp.roch.mn.charter.com] has
quit [Remote host closed the connection]
07:15 < quaid> dcaro: re: network infrastructure, we'll have to ask
Kevin on list for more details
07:16 < quaid> I think we did talk about a private VLAN
07:16 < quaid> and the need to control our own DHCP
07:16 < ewoud> jclift: I haven't done so recently, but I think that
should work; are you sure HEAD is multiple commits?
07:16 -!- abaron [~abaron at 87.68.20.33.cable.012.net.il] has joined #ovirt
07:16 < jclift> ewoud: checking
07:16 -!- lagarcia__ [~lagarcia at 201.82.143.114] has joined #ovirt
07:16 -!- mpastern [~mpastern at 94.159.206.219] has joined #ovirt
07:17 < ewoud> quaid: I'm not sure I like to separate the DB from the
app unless it's really needed
07:17 -!- mkolesni_ [~mkolesni at nat-pool-tlv-t1.redhat.com] has quit
[Quit: Bye!]
07:18 < ewoud> only listening on localhost makes security a bit easier imho
07:18 < dcaro> quaid: ok, we want to set up foreman with provisioning
right?
07:18 < ewoud> if possible
07:18 < quaid> ewoud: fair point about DBs, as long as it's not a
resource problem on the host, right?
07:18 < quaid> dcaro: aiui, yes
07:19 -!- ykaul [~ykaul at nat-pool-tlv-t1.redhat.com] has joined #ovirt
07:19 < quaid> #action Team needs to start writing up initial Puppet
manifests
07:19 < ewoud> quaid: that as wel; splitting it into two different VMs
on the same host gives little benefit imho
07:19 < quaid> ewoud: heh, true
07:19 -!- eedri_ [~eedri at nat-pool-tlv-u1.redhat.com] has joined #ovirt
07:19 < quaid> eedri_: we're actually meeting, the bot is down
07:19 < ewoud> only thing that might be easier is backups; one central
DB may be easier to back up
07:20 < quaid> eedri_:
http://etherpad.ovirt.org/p/new_hosting_design_Jan_2013
07:20 < dcaro> I've started writting the jenkins master/slave manifests
07:20 -!- mkolesni [~mkolesni at nat-pool-tlv-t1.redhat.com] has joined #ovirt
07:20 < eedri_> quaid, ok
07:20 < ewoud> dcaro: have you checked out puppet-jenkins?
07:20 -!- msalem [~msalem at nat-pool-tlv-t1.redhat.com] has quit [Read
error: Connection reset by peer]
07:20 < ewoud> dcaro: https://github.com/rtyler/puppet-jenkins
07:20 < jclift> ewoud: http://fpaste.org/z308/
07:20 < jclift> ewoud: Looks like multiple commits to me?
07:20 < dcaro> ewoud, yep, I'm using it as a base
07:21 < ewoud> jclift: that's log origin/master..HEAD or something
similar I assume
07:21 < quaid> ewoud: I was thinking about that, figured if we're doing
e.g. sql dumps, those are automatable - the challenge would be perhaps
more in mirroring databases if we had many of them
07:21 < jclift> ewoud: Yep
07:21 < quaid> I'm forgetting, did we setup the git repo for Infra?
07:21 < ewoud> dcaro: can I see the work somewhere?
07:22 < quaid> how does it come in to play here? does foreman/puppet
want all our configs in packages, etc.?
07:22 -!- emesika [~emesika at 46.120.148.109] has quit [Read error:
Connection timed out]
07:22 < ewoud> quaid: no, you can install files either using straight
content or through a template
07:22 -!- emesika [~emesika at 46.120.148.109] has joined #ovirt
07:22 -!- lagarcia_ [~lagarcia at 201.82.143.114] has quit [Ping timeout:
480 seconds]
07:23 < dcaro> ewoud: I'll push it to my fork as soon as I have
something worth looking into
07:23 < ewoud> jclift: which commit? I'd assume the first because the
last will need the others
07:23 < ewoud> dcaro: cool
07:23 -!- movciari [~movciari at nat-pool-brq-t.redhat.com] has joined #ovirt
07:24 < jclift> ewoud: All of mine in that list.  7 in total.
07:24  * jclift notices there's a GitHub mirror for all of the oVirt code
07:24 < ewoud> jclift: yes, but can you paste the output of git push? it
should mention the review requests
07:24 < jclift> ewoud: Can I just to a GitHub pull request instead of
this Gerrit crap?
07:24 < jclift> :)
07:25 < ewoud> jclift: no, they're meant as a read-only method
07:25 < jclift> Damn.  This is _way_ more difficult than it needs to be. :(
07:25 -!- doron [~doron at nat-pool-tlv-u1.redhat.com] has quit [Remote
host closed the connection]
07:26 < quaid> jclift: are you having the problem with gerrit that's
being reported on infra@?
07:26 -!- doron [~doron at nat-pool-tlv-u1.redhat.com] has joined #ovirt
07:26 < quaid>
http://lists.ovirt.org/pipermail/infra/2013-January/001785.html
07:26  * jclift looks
07:26 < quaid>
http://lists.ovirt.org/pipermail/infra/2013-January/001771.html
07:26 < jclift> ewoud: http://fpaste.org/gTZY/
07:26 < jclift> quaid: No.
07:27 < quaid> jclift: ok, didn't sound like it but ...
07:27 < jclift> jclift: I think I'm just having conceptual issues with
Gerrit. :)
07:27 < ewoud> jclift: that looks very weird; maybe HEAD isn't the
branch you want to push?
07:27 < jclift> After this patch, I'm never touching it again.
07:27 < jclift> ewoud: master HEAD is where I have the patches.
07:28 < jclift> I first tried with a branch "typo_fixes_v2", but wasn't
sure how to do stuff with branches with Gerrit.
07:28 < jclift> So, I just did it in master instead.
07:28 < ewoud> jclift: if you do branches I think git-review makes your
life a lot easier
07:28 -!- eedri_ [~eedri at nat-pool-tlv-u1.redhat.com] has quit [Ping
timeout: 480 seconds]
07:29 < jclift> I have no intention of ever doing anything with the
oVirt source ever again.
07:29 < jclift> (unless it moves to GitHub or something sane)
07:29 < jclift> So, not really going to invest time into learning new
tools just for this purpose.
07:29 < jclift> Hmmm.
07:29 -!- lvernia [~lvernia at nat-pool-tlv-t1.redhat.com] has quit [Quit:
Konversation terminated!]
07:30 < jclift> I think I'll just "abandon" this change set, and someone
else who knows Gerrit can fix the typos in things when they're bothered. :)
07:30 < ewoud> jclift: http://gerrit.ovirt.org/#/c/5541/ is the only
commit that shows up?
07:30 < jclift> ewoud: Yep
07:31 < ewoud> jclift: you could try to push to
refs/for/master/typo_fixes to make a branch
07:32 < ewoud> jclift: btw, I think I see what's the problem
07:32 < ewoud> all the commits have the exact same change id
07:32 < jclift> Yep
07:32 < ewoud> which means gerrit things it's the same change
07:32 < jclift> That ties them together doesn't it?
07:32 < quaid> btw, who is writing the master/slave hostgroups in to the
Etherpad?
07:32 < jclift> They're part of the same change.
07:32 < ewoud> jclift: no, a change ID means it's a unique trackable commit
07:33 < jclift> Hmmm.  I don't have an "Abandon Change" button.
07:33 < jclift> How do I abandon/nuke/etc this change?
07:33 < ewoud> jclift: just give the other commits a new ID, push and it
should work
07:33 < ewoud> jclift: I'll assume the first commit is still valid
07:34 < ewoud> jclift: change IDs are a means of tracking multiple
versions of a single patch
07:34 < jclift> How do I tie multiple patches into one related "change" ?
07:34 < jclift> Or is that not possible?
07:34 < ewoud> jclift: I think you push to refs/for/master/<my_branch_name>
07:34 < jclift> k
07:35 < quaid> ah, it's eedri
07:35 -!- jbrooks [~jason at c-50-148-148-82.hsd1.ca.comcast.net] has
joined #ovirt
07:35 -!- eedri [~eedri at nat-pool-tlv-t1.redhat.com] has joined #ovirt
07:37 < quaid> ewoud: regarding the smart proxy, is that something we'd
want bundled on the same server with Puppet, or would that be a
stand-alone host, too?
07:38 < dneary> quaid, What agenda item are we on? I'm a little lost
07:38 < ewoud> quaid: We need to run puppet on the smartproxy because it
managed /etc/puppet/autosign.conf for example
07:38 < dneary> Ah - hosting plan
07:38 < dneary> Never mind me
07:39 < quaid> yep, and on the Etherpad
07:39 < quaid> http://etherpad.ovirt.org/p/new_hosting_design_Jan_2013
07:39 < quaid> ewoud: ah, ok
07:39 -!- shireesh [~shireesh at nat-pool-sin2-t.redhat.com] has joined #ovirt
07:39 < ewoud> also, we need to install puppet from the puppetlabs repo
because the server needs to be >= clients
07:40 < ewoud> and afaik F17 is puppet 2.7 where EPEL6 is 2.6
07:40 < eedri> ewoud, i think we have puppet master on rhel and it
managed fedora clients
07:41 < ewoud> eedri: I have no fedora clients, but pretty sure I heard
something about that
07:41 -!- kamdard [~kamdard at pool-96-226-22-78.dllstx.fios.verizon.net]
has joined #ovirt
07:41 < ewoud> eedri: we can start with EPEL6 version and upgrade if it
doesn't work
07:41 -!- ofri_ [~ofri at nat-pool-tlv-u1.redhat.com] has quit [Ping
timeout: 480 seconds]
07:41 < eedri> ewoud, sure
07:41 -!- oliel [~oliel at nat-pool-tlv-t1.redhat.com] has quit [Remote
host closed the connection]
07:41 < ewoud> eedri:
http://etherpad.ovirt.org/p/new_hosting_design_Jan_2013 btw
07:41 < dcaro> Puppet is sensitive to the versions of the master/client,
we should use the same on clinet and master
07:43 -!- achub [~achub at nat-pool-tlv-t1.redhat.com] has quit [Quit: Leaving]
07:44 -!- ofri_ [~ofri at nat-pool-tlv-t1.redhat.com] has joined #ovirt
07:44 < quaid> hmm, well I reckon if it works it will work fine, and if
it doesn't, it will be clear it's not working? (Having EPEL's Puppet run
Fedora slaves)
07:44 < ewoud> jclift: any luck changing the IDs?
07:45 < eedri> ewoud, are you chatting on the pad? cause i don't see it
07:45 < eedri> ewoud, or just reviewing the items
07:45 < ewoud> eedri: we're reviewing the items
07:45 -!- gchaplik [~gchaplik at nat-pool-tlv-t1.redhat.com] has joined #ovirt
07:45 < jclift> ewoud: I'm just about to try this:
07:45 < jclift> git push gerrit.ovirt.org:ovirt-engine
typo_fixes_v3:refs/for/master/typo_fixes_v3
07:45 < jclift> Does that sound right?
07:46 < ewoud> jclift: I think it does
07:46 < jclift> The new commits, with changed id's, are in a new
"typo_fixes_v3" branch.
07:46 < jclift> k, lets see what happens.
07:46 < quaid> we're chatting here because I'll paste the logs to the
mailing lists for archiving when we're done
07:46 -!- ljk
[~ljk at burl-mse-71-255-129-12.static.ngn.east.myfairpoint.net] has joined
#ovirt
07:47 < ewoud> jclift: looks like it pushed multiple commits now
07:47 < eedri> quaid, btw, any update on gerrit issue?
07:47 < eedri> quaid, seems issue pressit from tlv site
07:47 < ewoud> eedri: does HTTP work normal? traceroue?
07:48 < quaid> eedri: are all the folks affected so far in the same office?
07:49 < eedri> ewoud, http from browsers works
07:49 < eedri> ewoud, yea, seems to all from tlv site
07:49 -!- sgrunewald [~soeren.gr at p5B3C048D.dip0.t-ipconnect.de] has quit
[Ping timeout: 480 seconds]
07:49 < eedri> ewoud, but it works for some
07:49 < ewoud> eedri: weird that http from browsers would work, but git
over http doesn't
07:49 < ewoud> eedri: to me it feels like you should contact your
network admin
07:50 < eedri> ewoud, ssh works.. but http/git doesn't
07:50 < eedri> ewoud, so i'm not sure it's a network issue
07:51 < jclift> ewoud: Is that good or bad?  It's now 7 separate changes.
07:51 < jclift> Should I just leave it alone now, or just abandon it
something.
07:51 < jclift> I'm completely over putting any more effort into this. :)
07:52 < ewoud> jclift: yes, every commit becomes a change and I think
you're fine now and just wait for a review
07:52 < jclift> ewoud: k, thanks.
07:52 -!- aglitke [~aglitke at 68-115-116-10.dhcp.roch.mn.charter.com] has
joined #ovirt
07:52 -!- jlibosva [~jlibosva at nat-pool-brq-t.redhat.com] has quit [Quit:
Leaving.]
07:52 < jclift> I really hope you guys move to GitHub or some "better
than Gerrit" system. :)
07:53 < ewoud> jclift: I'm not sure how easy GH is to manage multiple
branches or from a release point of view
07:54 < ewoud> quaid: eedri dcaro I tried to make a timeline at the end
of the document, got anything to add?
07:54 < jclift> Every other project on the planet seems to manage it. ;)
07:54  * quaid looks
07:55 < dcaro> should we install ovirt to handle the vms?
07:56 < ewoud> dcaro: we've talked about it a bit, but never reached a
conclusion
07:56 < eedri> ewoud, don't we have foreman server already?
07:56 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has quit [Read
error: Connection reset by peer]
07:56 < ewoud> eedri: I set up a PoC on a VM of mine, but I think we'd
prefer to run it all on the new alterway machines
07:56 < quaid> eedri: so ... the nodes that are masters in the
Puppet/Foreman arrangement are *not* provisioned using themselves?
07:56 < eedri> ewoud, sure
07:56 < quaid> i.e., puppet, foreman, smartproxy
07:56 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has joined #ovirt
07:57 < eedri> quaid, its the chicken and egg
07:57 -!- chouchins [~chouchins at IP-214-60.wustl.edu] has joined #ovirt
07:57 < eedri> quaid, but ewoud and david can explain better than i
probably
07:57 < quaid> I've got a phone call in a few minutes, I think ... but
I'm definitely still working on the Etherpad & can keep in this meeting
07:57 < eedri> quaid, what's the best approaach
07:57  * quaid will work on the bot later today
07:57 < eedri> i think installing foreman + puppet master from stable
rpms will be best
07:58 -!- sgrunewald [~soeren.gr at p5B3C028F.dip0.t-ipconnect.de] has
joined #ovirt
07:58 < quaid> +1
07:58 < dcaro> +1
07:58 < eedri> then using existing install to manage and provision other
servers
07:58 < ewoud> eedri: foreman 1.1 is about to be released, but maybe
ohadlevy can tell us more
07:58 < quaid> I want to use-or-fix what's in EPEL, if we can
07:58 < dneary> Stepping away... quaid please ping me if you need my
input on anything - I think I already updated infra tickets related to me
07:58 < quaid> dneary: thanks
07:58 < eedri> quaid, when can we get root access to alterway1?
07:59 < quaid> yeah, if anyone wants to review their own Trac tickets
and update them, we can skip that part of this meeting
07:59 < quaid> eedri: Kevin said he's installing base OS today
07:59 < eedri> quaid, ok.
07:59 < eedri> quaid, assuming we don't have foreman server yet, we'll
probably need to install jenkins master manually for no
07:59 < eedri> w
07:59 -!- tjelinek [~tjelinek at nat-pool-brq-t.redhat.com] has quit [Read
error: Operation timed out]
07:59 < quaid> ok
08:00 < ewoud> dcaro: I think running ovirt can be good with eating your
own dogfood, but not sure how much overhead it is when we only have one
machine running VMs
08:00 < quaid> I think it's OK to push ahead in parallel as long as we
are including the manual installs in Foreman later
08:01 < eedri> quaid, we can later on add the relevant hostgroups to
manage already installed servers
08:01 < quaid> great
08:01 < quaid> *whew*
08:02 < dcaro> ewoud: we have at least two right? alterway02 and
rackspace01? That's what's on the pad
08:03 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has quit [Quit:
Leaving.]
08:03 < ewoud> dcaro: yes, but can we manage those two over a WAN link?
08:03 < ale_l> out of curiosity clikced that link :
http://etherpad.ovirt.org/p/new_hosting_design_Jan_2013 - is it supposed
to be public ?
08:03 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has joined #ovirt
08:03 < ewoud> ale_l: yes, nothing really secret I think
08:03 < ale_l> ah ok :)
08:04 -!- mkrcmari [~mkrcmari at nat-pool-brq-u.redhat.com] has quit [Quit:
Leaving]
08:04 -!- aglitke [~aglitke at 68-115-116-10.dhcp.roch.mn.charter.com] has
quit [Remote host closed the connection]
08:06 < quaid> ale_l: thanks for checking, we only keep actual passwords
secret, otherwise we're radically transparent infrastructure :)
08:07 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has quit []
08:08 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has joined #ovirt
08:09 -!- ljk
[~ljk at burl-mse-71-255-129-12.static.ngn.east.myfairpoint.net] has quit
[Remote host closed the connection]
08:09 -!- masayag [~masayag at nat-pool-tlv-t1.redhat.com] has quit [Remote
host closed the connection]
08:09 -!- aglitke [~aglitke at 68-115-116-10.dhcp.roch.mn.charter.com] has
joined #ovirt
08:10 -!- mkrcmari [~mkrcmari at nat-pool-brq-t.redhat.com] has joined #ovirt
08:10 -!- masayag [~masayag at nat-pool-tlv-t1.redhat.com] has joined #ovirt
08:11 -!- alourie [~alourie at nat-pool-tlv-t1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:12 -!- mkletzan [~mkletzan at nat-pool-brq-t.redhat.com] has quit [Quit:
BRB]
08:12 -!- alonbl_ [~alonbl at 0001aebf.user.oftc.net] has quit [Ping
timeout: 480 seconds]
08:13 < quaid> ewoud: dcaro I've been pondering the usage of oVirt
itself, and I think perhaps we'll find it worthwhile, but I think we're
OK proceeding without oVirt as a layer & adding it layer
08:13 -!- ohadlevy [~ohadlevy at nat-pool-tlv-t1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:14 -!- mskrivanek is now known as mskrivanek|away
08:15 < dcaro> ewoud, mmm, not sure how much bandwith should it take but
I think it should not be too much right?
08:16 -!- tjikkun_work [~tjikkun at 2a00:d10:2:1:a6ba:dbff:fef3:f431] has
quit [Quit: Ex-Chat]
08:16 -!- mpastern [~mpastern at 94.159.206.219] has quit [Read error:
Operation timed out]
08:17 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has quit [Remote
host closed the connection]
08:18 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has joined #ovirt
08:18 < dcaro> quaid, I think that the benefits of having ovirt will
start to become relaly noticeable with the migration, as we are starting
to have more vms (also I think that using our own software should be a
must xd)
08:19 < quaid> I'm +1 with doing it, for sure
08:19 -!- mkolesni [~mkolesni at nat-pool-tlv-t1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:19 < quaid> if we can do it as part of the migration, even better
08:27 -!- aglitke [~aglitke at 68-115-116-10.dhcp.roch.mn.charter.com] has
quit [Read error: Connection timed out]
08:27 -!- danken [~danken at bzq-109-67-31-252.red.bezeqint.net] has quit
[Quit: Leaving.]
08:27 -!- aglitke [~aglitke at 68-115-116-10.dhcp.roch.mn.charter.com] has
joined #ovirt
08:28 -!- mestery [~mestery at 128-107-239-233.cisco.com] has quit [Remote
host closed the connection]
08:28 -!- ade [~abradshaw at 109.58.116.227.bredband.tre.se] has quit [Ping
timeout: 480 seconds]
08:28 < ewoud> quaid: I think it mostly starts to make sense if we add
more than one physical machine
08:28 -!- crobinso [~crobinso at cpe-67-248-219-87.nycap.res.rr.com] has
joined #ovirt
08:29 < ewoud> btw, what's the reason we need to use alterway02 as a
dedicated machine?
08:29 -!- lhornyak [~lhornyak at nat-pool-brq-u.redhat.com] has quit [Read
error: Operation timed out]
08:29 -!- rvaknin [~rvaknin at nat-pool-tlv-t1.redhat.com] has quit [Read
error: Operation timed out]
08:31 -!- aglitke [~aglitke at 68-115-116-10.dhcp.roch.mn.charter.com] has
quit [Remote host closed the connection]
08:32 < quaid> ewoud: for Jenkins?
08:32 < quaid> aiui, we want Jenkins on it's own host, but why did I get
that?
08:32 < quaid> * Read we don't want to have slaves on the same host
08:32 -!- anande [~anande at nat-pool-pnq-t.redhat.com] has joined #ovirt
08:33 < ewoud> quaid: yes, I think eedri said something like that but
I'm wondering if we get sufficient performance from a VM
08:33 < ewoud> if a VM is sufficient we could go for ovirt
08:33 < quaid> I guess the need for nested qemu v. bare metal (iirc) is
more on the slave side?
08:34 < ewoud> yes
08:34 < quaid> ah
08:34 -!- djasa [~djasa at nat-pool-brq-t.redhat.com] has quit [Quit: Šmitec]
08:34 -!- shireesh [~shireesh at nat-pool-sin2-t.redhat.com] has quit [Ping
timeout: 480 seconds]
08:34 < ewoud> in the ideal situation the master is only a master and
doesn't run jobs afaik
08:35 < quaid> ewoud: btw & OT, I'm almost definitely coming to Brussels
on Friday night & spending most of Saturday at FOSDEM
08:35 < quaid> ok, no problems, master doesn't run jobs, but yeah, if it
can be a VM we can use that host more efficiently
08:35 < ewoud> quaid: cool
08:35 < eedri> quaid, ewoud i've also tried to think if jenkins really
needs it's own single bare metal server
08:35 < ewoud> quaid: I'm not sure what jenkins does under higher load
of many jobs
08:36 < eedri> quaid, ewoud but i can't know for sure if running it on a
vm will provide enough performance
08:36 < ewoud> quaid: I think foreman devs are planning something on
saturday evening, like dinner
08:36 < quaid> RobertM: did you have some insight about Jenkins on its
own bare metal v. VM?
08:36 < eedri> quaid, perhaps the bad experience we had from using an
amazon vm affects the decision
08:37 < quaid> ewoud: cool, I'm heading back to Paris sometime on
Saturday night, but it could be a late train ride
08:37 < eedri> quaid, ewoud i can say we're using vms to run
foreman,puppet,gerrit
08:38 -!- ppinatti [~paulo at 187.74.69.163] has quit [Quit: Leaving.]
08:38 < ewoud> we run our jenkins master at $employer virtual, but we
have very little jobs
08:38 -!- ppinatti [~paulo at 187.74.69.163] has joined #ovirt
08:38 < eedri> ewoud, we have lots of jobs..
08:38 -!- alkaplan [~alkaplan at nat-pool-tlv-t1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:38 <@RobertM> quaid, I can tell that the master uses a lot of IO so
what every you go with needs to strong IO.
08:39 < ewoud> eedri: that's what makes my experience harder to extrapolate
08:39 -!- Udayendu [~ukar at nat-pool-pnq-t.redhat.com] has joined #ovirt
08:39 < ewoud> RobertM: was the master running jobs as well?
08:39 < quaid> RobertM: I think we have that included, though I haven't
seen the actual specs from AlterWay, we included fast IO in the initial
requests
08:41 <@RobertM> I was running jobs at first but as the number of slaves
went up the number of jobs I could run on the master went down.
08:42 -!- ovedo [~oourfali at nat-pool-tlv-t1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:43 -!- alourie [~alourie at 80.250.158.201] has joined #ovirt
08:43 -!- shireesh [~shireesh at nat-pool-blr-t.redhat.com] has joined #ovirt
08:43 <@RobertM> and that was with a sata drive
08:45 -!- chouchins [~chouchins at IP-214-60.wustl.edu] has quit [Remote
host closed the connection]
08:45 <@RobertM> Most VM's dont have enough IO to handle the load on the
master
08:45 -!- lhornyak [~lhornyak at nat-pool-brq-t.redhat.com] has joined #ovirt
08:46 < quaid> so I'm unsure now, was the problem that you were also
running jobs on the master? or is it that ultimately Jenkins master will
collapse on a VM and needs bare metal with fast IO?
08:46 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has quit [Quit:
Leaving.]
08:47 -!- mbetak [~mbetak at nat-pool-brq-t.redhat.com] has joined #ovirt
08:47 < eedri> RobertM, but if we assign an external io to the master,
fast enough.. it might be enough.
08:48 <@RobertM> quaid, Ever job requires you to copy to the slave then
copy back from the slave that is a lot of IO the more jobs you run the
higher that IO gets.  Please that becomes a bottle neck for the entire
process.  So do skimp on the master
08:48 <@RobertM> please=plus
08:49 <@RobertM> do=dont
08:49 < quaid> heh, that wsa a useful typo fix, thanks
08:49 <@RobertM> Also my keyboard is failing so it is really effecting
my ability to type.
08:50 < quaid> I'd say, we could probably start with Jenkins owning that
host, and watch the usage over time - if it ends up having spare CPU &
RAM, we could consider changing and watching
08:50 -!- alonl [~alon at nat-pool-tlv-u1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:50 < eedri> quaid, sounds good
08:50 < eedri> quaid, if we'll see we're short on bare metal hosts
08:50 < eedri> quaid, it will be easy enough to install a new jenkins vm
on existing system
08:53 -!- masayag [~masayag at nat-pool-tlv-t1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:54 -!- |AbsyntH|
[~never at host250-66-static.36-85-b.business.telecomitalia.it] has quit
[Quit: I'm an apatheist. The question is no longer interesting, and the
answer no longer matters]
08:56 -!- ofri_ [~ofri at nat-pool-tlv-t1.redhat.com] has quit [Ping
timeout: 480 seconds]
08:56 < quaid> +1
08:56 <@RobertM> CPU and Ram requirements are pretty minor.  My system
has 16G or ram and 8 core and it was way overkill.  I found the master
only used around 6G or ram and the CPU were never busy.  Now IO on the
sata drive got nuts.
08:56 -!- ykaplan [~ykaplan at nat-pool-tlv-t1.redhat.com] has quit [Remote
host closed the connection]
08:59  * eedri has to go
08:59 < eedri> i'll update the pad with more topics if i'll think on
something

## end IRC log
-- 
Karsten 'quaid' Wade, Sr. Analyst - Community Growth
http://TheOpenSourceWay.org  .^\  http://community.redhat.com
@quaid (identi.ca/twitter/IRC)  \v'  gpg: AD0E0C41

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 253 bytes
Desc: OpenPGP digital signature
URL: <http://lists.ovirt.org/pipermail/infra/attachments/20130109/aff9458c/attachment.sig>


More information about the Infra mailing list