<desai> 00:01:10 actually you should be able to pull directly from https://svn.mcs.anl.gov/repos/bcfg/trunk/bcfg2/src/sbin/bcfg2-reports
00:01:17 <desai> no backporting was actually needed iirc
00:02:44 <joe> desai, cool. thanks. I have a working copy of trunk on this system already...
00:11:20 <joe> heh, ./bcfg2-reports (from trunk) bitches as such: Failed to load configuration settings.is /etc/bcfg2.conf readable? when it is -rw-r--r-- root root
00:12:06 <desai> do you have the db settings in there?
00:13:08 <joe> it's all default to sqlite3 stuff
00:13:45 <joe> I plan on using mysql or postgres but not yet
00:14:14 <kisielk> I found I was getting that error from database misconfiguration in the past
00:14:20 <joe> http://pastebin.com/m19a625b
00:14:28 <kisielk> ah, I was just about to ask
00:14:29 <joe> all as they were w/ the initial conf file
00:15:29 <kisielk> hmm
00:15:44 <kisielk> can you pastebin your whole config?
00:17:24 <joe> sure one sec
00:18:45 <joe> http://pastebin.com/m41a9ada9
00:20:16 <kisielk> add plugins = DBStats
00:20:18 <kisielk> under [server]
00:20:21 <kisielk> then restart the server
00:20:54 <joe> cool, thanks
00:23:23 <desai> did that do the trick?
00:23:26 <joe> now that's odd I ran bcfg2 -I and it didn't prompt me to update /etc/bcfg2.conf wtf?
00:24:20 <desai> is /etc/bcfg2.conf wrong?
00:25:49 <joe> desai, nope it updated it
00:27:28 <desai> wait a sec
00:28:00 <desai> so you ran the client with -I (capital I) and it didn't prompt you?
00:28:10 <desai> is this a backrevved client, or 0.9.6?
00:28:29 <joe> it's just 0.9.5.7
00:28:40 <desai> that is pretty well-tested at this point
00:28:45 <joe> I would imagine
00:28:52 <joe> found it rather strange
00:28:53 <desai> are you sure that bcfg2.conf was updated at that time?
00:29:09 <desai> -I behavior is pretty well hardwired in to prompt
00:29:17 <desai> which options did you run with?
00:29:28 <joe> bcfg2 -I that's it
00:29:34 <desai> weird
00:29:38 <joe> this is the server and a client obviously
00:29:42 <desai> right
00:29:56 <desai> hm
00:29:58 <desai> wait a sec
00:30:24 <desai> nm, that isn't it
00:30:31 <joe> hehe
00:30:45 <desai> can you try making another change to bcfg2.conf on the server, and run the client again with -v -d -I
00:32:15 <joe> <sigh> I'm a dork I edited the wrong one ....
00:32:25 <joe> sign I need more caffeine
00:32:29 <joe> sorry
00:32:47 <desai> no problem
00:32:58 <joe> I edited the actual /etc/bcfg2.conf by mistake
00:33:00 <desai> ACTION breathes a sigh of relief ;)
00:33:03 <joe> silly context
00:33:27 <joe> ACTION thought you just needed a bit of adrenaline rush ;)
00:42:54 <kisielk> I'm really starting to like SLES
00:43:05 <kisielk> other than the fact it doesn't have Python 2.5 in the current release
00:43:47 <kisielk> and I haven't yet figured out how to make my own package repository
00:49:53 <joe> keep drinking the koolaid ;P
00:52:55 <kisielk> mmm koolaid
00:53:01 <kisielk> actually, I hate koolaid
00:53:43 <dopp> kisielk: you can use YUM repos
00:53:51 <dopp> yast/zyper/rug will all read YUM
00:53:58 <kisielk> dopp: yes, but how do I create a repository ?
00:54:07 <kisielk> ideally I want to mirror all the official packages, and add my own
00:54:18 <dopp> with createrepo... just like you'd do on redhat
00:54:24 <kisielk> hm ok
00:54:29 <kisielk> does createrepo work on gentoo ? :)
00:54:46 <dopp> couldn't tell ya
00:54:52 <dopp> a tool for making YUM repos
00:55:01 <kisielk> okay, I'll look in to that
00:55:02 <joe> kisielk, hehe, no
00:55:20 <kisielk> well, our software repository is a Gentoo server
00:55:20 <dopp> if you want to make a real YaST repo, then you need to do some research
00:55:21 <joe> kisielk, actually it might be able to
00:55:24 <kisielk> so it would be nice to have all the RPMs there
00:55:30 <dopp> I've done it, though I can't recall the name of the commands off the top of my head
00:55:34 <dopp> it's been awhile
00:55:51 <kisielk> it actually only needs to be good enough for RPMng to work with it
00:55:58 <kisielk> I'm not sure what kind of repository it needs
00:56:06 <dopp> yeah, I doubt RPMng can use YaST repos anyway
00:56:29 <dopp> kisielk: one'd hope RPMng has docs ;)
00:56:42 <dopp> I'll likely not have RPMng manage all packages in the OS
00:56:47 <dopp> only ones that I 'bless'
00:57:00 <dopp> ACTION prefers YaST for package management anyway
00:57:03 <kisielk> well, I plan to put the whole base install in to Base
00:57:08 <kisielk> then prune away the crap I don't want
00:57:12 <dopp> but almost all of my clients are diskless, heh
00:57:32 <kisielk> ideally I want to be able to make new machines by just doing a minimal install
00:57:34 <kisielk> then running bcfg2
00:57:35 <kisielk> and done
00:57:38 <dopp> you may run into issues with SLES updates if you start using YUM repos
00:57:59 <dopp> Novell uses a thing they call 'patches'. They're RPMs + a number of scripts
00:58:12 <dopp> yast, zyper and rug are the only things I've found that deal with them appropriately.
00:58:20 <kisielk> hm interesting
00:58:30 <kisielk> and they don't make the full package available?
00:58:32 <dopp> well, more correctly, they're RPMs, patch.rpms, delta.rpms, and a number of scripts
00:58:49 <dopp> kisielk: they do, but all of the other little bits keep things clean
00:59:10 <dopp> we found that over the course of, say, a year, a strictly YUM-managed SLES system suffered some bad cruft
00:59:11 <kisielk> how do they expect you to update 100 machines?
00:59:15 <kisielk> using zvm ?
00:59:28 <dopp> kisielk: either zmd + rug, or with zypper
00:59:38 <kisielk> hm that kind of sucks
00:59:46 <dopp> or do what I do and use YaST on 1 machine, take an image and push the image out to hundreds of machines ;)
00:59:57 <kisielk> well, yeah that works for that case
01:00:01 <dopp> though, that's left as an exercise to the customer ;)
01:00:06 <kisielk> but that only covers about half the machines on our network
01:00:06 <desai> sounds like we should write a zypper driver for bcfg2
01:00:19 <dopp> desai: Yeah, I mentioned that to you at SC
01:00:33 <dopp> the thing is, zyper isn't 100% reliably scriptable
01:00:49 <dopp> they've atleast added the ability to explicitly and permanently trust unsigned packages
01:00:54 <dopp> otherwise it would prompt you EVERY TIME
01:00:57 <dopp> it was maddening
01:01:08 <kisielk> zyper is not in the default install?
01:01:30 <dopp> kisielk: it's in the base OS... I dunno what you define as the 'default install'
01:01:42 <kisielk> oh, it's zypper
01:01:44 <dopp> it's likely not in the Base package pattern, if that's what you mean
01:01:52 <dopp> yes, sorry, my typo
01:02:20 <desai> dopp: yeah, that rings a bell now. sorry, i was pretty exhausted by the time i made it to austin ;)
01:02:54 <dopp> when SLES10 landed, you had two choices: YaST or rug. YaST isn't scriptable at all, and rug is really ghastly (but it can do things like roll back your OS to a point in time, which is neat)
01:02:54 <kisielk> all this stuff is so complicated heh
01:03:14 <kisielk> sometimes I just start trying to think of the big picture
01:03:17 <kisielk> and my head nearly explodes
01:03:25 <dopp> ZMD (the MONO-based monstrosity that rug talks to) uses a database (or the like) called zypp
01:03:44 <dopp> so some SuSE devs got together and wrote zypper to modify the ZMD database without having to go through ZMD
01:03:52 <dopp> ZMD takes like 500MB of mem... it's absurd
01:03:59 <kisielk> wow
01:04:00 <kisielk> heh
01:04:04 <kisielk> go go mono
01:04:39 <dopp> we chkconfig novell-zmd off on all our systems and have some home-grown scripts wrapped around zypper
01:04:56 <dopp> atleast that's on the SLES systems that I don't manage (central mail, web, SLED workstations, etc, etc)
01:05:22 <dopp> fortuantely, I don't need to use zypper on my clusters and Altii 'cause the systems with disks are so few
01:05:32 <kisielk> yes
01:05:38 <kisielk> so you load your whole system image in to RMA?
01:05:40 <kisielk> er, RAM?
01:05:42 <dopp> yup
01:05:49 <kisielk> how big is it? and how much RAM do you have?
01:06:06 <dopp> it's 600MB to 1.1GB, depending on the cluster.
01:06:06 <kisielk> I was actually thinking about doing that, so we're not dependent on our NFS root, since it's a pretty nasty single point of failur
01:06:13 <dopp> all our systems have 2GB/core
01:06:19 <kisielk> but our nodes only have 2GB / RAM
01:06:20 <dopp> so 8-32 GB
01:06:23 <dopp> yeesh
01:06:34 <dopp> yeah, we fortunately don't need nfs root
01:06:35 <kisielk> well, our application is not very RAM intensive
01:06:49 <kisielk> but I've been thinking of getting solid state disks
01:07:00 <kisielk> to host the OS image
01:07:07 <dopp> we actually did roll out warewulf with NFS root on one cluster, but then changed to diskless 'cause bringing the cluster up took too long
01:07:17 <kisielk> you can get them for ~$100 now a days
01:07:19 <kisielk> for a few gigs
01:07:23 <kisielk> which would be plenty
01:07:26 <dopp> one that's way to go
01:07:39 <dopp> we use all blades, for the most part... not everything even has a USB port ;)
01:07:51 <kisielk> well, they're SATA
01:08:22 <kisielk> but we use pizza boxes, our cluster is colocated
01:08:22 <dopp> we have disks in most of our nodes for local scratch, but we just don't use them for the OS
01:08:34 <dopp> OS + scratch makes the node bog down when a job starts hitting the disk
01:08:44 <kisielk> so I don't want to have to get any fancy power setup from the host
01:08:50 <dopp> yeah
01:08:52 <kisielk> since they'll probably charge an arm and a leg for that
01:09:06 <dopp> ACTION demands 60A 3-phase :)
01:09:14 <dopp> and a lot of it
01:09:15 <kisielk> and the Altix XE's are pretty dense
01:09:21 <kisielk> almost too dense heh
01:10:00 <dopp> yeah
01:10:04 <kisielk> they're something like 6A each, which is a lot for a 1U box
01:10:08 <dopp> I've started getting various mobo failures in mine
01:10:18 <kisielk> in the 310's ?
01:10:24 <dopp> kisielk: oh, yeah, we had to do a lot of power balancing when we got our 5 racks of XEs
01:10:28 <dopp> kisielk: yeah
01:10:38 <kisielk> yeah out of the 6 we got
01:10:45 <kisielk> 4/12 units had bad BMCs
01:10:49 <dopp> NICs dying, HCAs dying, even some USB buses causing panicks
01:10:59 <dopp> oh, BMC failures are a monthly occurance
01:11:03 <kisielk> heh
01:11:12 <dopp> the supermicro BMCs are crap
01:11:26 <kisielk> everyone's BMCs are crap it seems
01:11:31 <kisielk> the IBM ones we have are even worse
01:11:41 <dopp> yeah, mostly. I've had decent luck with intels (like in the SGI XE240)
01:11:54 <kisielk> sometimes they forget their configuration
01:11:57 <dopp> and Dell's don't seem failure prone, though they lack a lot of features
01:12:03 <kisielk> or just stop working
01:12:10 <kisielk> until you pull the power and let the caps in the power supply drain
01:12:11 <dopp> yeah, lots of BMC shyness ;)
01:12:15 <dopp> kisielk: exactly!
01:12:22 <dopp> that's what we need to do in our XE310s
01:12:27 <kisielk> which kind of
01:12:30 <kisielk> entirely eliminates the point
01:12:31 <kisielk> of having a BMC
01:12:33 <dopp> dozens of them every time we do maintenance, heh
01:12:40 <kisielk> since the whole use for me
01:12:47 <kisielk> is so I don't have to bus down to the colo facility
01:12:50 <kisielk> to reboot the machine
01:12:50 <dopp> indeed
01:13:01 <dopp> fortunately, I only need to go down 5 stories ;)
01:13:07 <dopp> and not 5 miles
01:13:11 <kisielk> and it sucks even more with the XE310s
01:13:19 <kisielk> since there's only one power cord for 2 machines
01:13:29 <dopp> we have 12 (6 chassis) SE320s and they seem a little better
01:15:08 <kisielk> anyway, time to go rock climbing
01:15:15 <kisielk> and then install our new SGI gear tomorrow
01:15:18 <kisielk> FINALLY got here
18:09:10 <kisielk_home> hm, just tried to install the bcfg2 client on another system but it complains it can't find xml.dom
18:09:14 <kisielk_home> going to see what's up with that
18:40:52 <kisielk_mbp> hm
18:41:06 <kisielk_mbp> apparently you need to install python-xml as well
18:41:12 <kisielk_mbp> doesn't seem to be a dependency of the package
18:41:26 <kisielk_mbp> dopp: ping
19:05:10 <desai> teknix: is there currently a network auto-installation mechanism for nexenta?
19:31:19 <dopp> kisielk_mbp: yes?
19:31:30 <kisielk_mbp> dopp: see above
19:31:34 <dopp> yeah, saw that
19:31:48 <kisielk_mbp> is it a missing dep? or something else wrong?
19:31:59 <dopp> you do need python-xml, I believe
19:32:16 <dopp> the spec file in SVN has little in the way of dependencies specified
19:32:34 <dopp> so I just went by the wiki when I installed bcfg2
19:33:07 <dopp> gimme a minute and I'll dig up my list of packages that satisfy the deps
19:37:21 <dopp> kisielk_mbp: ok, here's what I have (12-line flood):
19:37:26 <dopp> Prereq Source
19:37:26 <dopp> -----------------------------------------------
19:37:26 <dopp> elementtree python-elementtree in SLE10 SDK
19:37:26 <dopp> pyopenssl python-openssl in SLE10 SDK
19:37:26 <dopp> libxml2 libxml2 in SLES10 base
19:37:29 <dopp> libxslt libxslt in SLES10 base
19:37:31 <dopp> fam fam-server in SLES10 base (only needed on bcfg2 server)
19:37:34 <dopp> lxml python-lxml-1.2.1-1 from ANL
19:37:36 <dopp> python-fam python-fam-1.1.1-1 from ANL
19:37:39 <dopp> python-distutils python-devel in SLES10 base (only required if building bcfg2 from source)
19:37:42 <dopp> python-rpm rpm-python in SLES10 base
19:37:45 <kisielk_mbp> yeah, I don't have a problem installing the stuff myself
19:37:54 <kisielk_mbp> I just mean that the RPM should have that info there
19:38:00 <dopp> indeed
19:38:07 <desai> if you have lxml, you don't need elementtree
19:38:29 <dopp> desai: so, if I have elementree can I get rid of lxml?
19:38:45 <kisielk_mbp> but yeah, we should add python-xml to the spec I guess, if it's not there
19:38:53 <dopp> ACTION would like to reduce the number of non-SLES packages as much as possible
19:39:06 <desai> dopp: there is a different story on the client vs the server
19:39:19 <desai> on the client, you need python 2.3+
19:39:21 <dopp> desai: yeah, I couldn't really find clear docs on it, heh
19:39:31 <desai> and elementtree or lxml
19:40:01 <desai> if you are running python 2.5+, a version of elementtree is included
19:40:08 <desai> so if you have python2.5, you are all set (for clients)
19:40:14 <dopp> right
19:40:22 <kisielk_mbp> yeah, but SLES still ships with 2.4
19:40:26 <desai> in any case, if you add m2crypto, things will go a _lot_ faster
19:40:39 <dopp> kisielk_mbp: but includes an elementtree package
19:40:46 <kisielk_mbp> yes
19:40:53 <desai> for the server side, you need lxml
19:40:57 <desai> it can't use elementtree
19:41:03 <dopp> desai: ok, thanks
19:41:10 <dopp> that can go as seperate requires in the spec file
19:41:21 <desai> that pyopenssl req is stale; we don't need it anymore
19:41:29 <desai> was that in the spec someplace?
19:41:36 <dopp> no, I think I found it in the wiki
19:41:44 <desai> ACTION goes to hunt it down
19:41:49 <dopp> there were two or three different pages with instructions
19:41:55 <dopp> wasn't sure which was up to date, if any
19:42:08 <dopp> so I just threww every package at it ;)
19:43:13 <desai> whoa, that fc5 quickstart needs some updatin'
19:43:19 <dopp> ok, so to make this clear, server needs python 2.3+ + lxml, clients need 2.3+ + elementtree (understanding that elementtree is included with 2.5) || python 2.3+ + lxml
19:43:56 <desai> server _needs_ python 2.3+, fam || gamin, lxml
19:44:10 <dopp> well, right, I just meant in terms of python + xml, heh
19:44:18 <bcfg2gozerbot> bcfg2-timeline: Prereqs edited by desai
19:44:27 <desai> clients need (python 2.3+, elementree || lxml)
19:44:49 <dopp> what about xslt and the like?
19:45:06 <desai> server-side only
19:45:09 <dopp> ok
19:45:45 <joe> any of you heard bad things about internap's cdn ?
19:48:30 <dopp> desai: where is m2crypto needed? client, server or both?
19:48:46 <desai> strictly speaking nowhere, but it is useful in both places
19:48:52 <dopp> ok, cool
19:48:59 <desai> like _really_ useful
19:49:04 <dopp> sadly, it's neither in the SLES 10 base nor in the SDK
19:49:04 <desai> ssl goes _slow_ without it
19:49:17 <dopp> ok, so it's m2crypto or pyopenssl?
19:49:21 <dopp> or both?
19:49:23 <desai> just m2
19:49:25 <dopp> ok
19:49:27 <desai> pyopenssl was an old req
19:49:30 <dopp> ok
19:49:54 <dopp> libxml2 needed on client and server?
19:49:56 <desai> i think that i've managed to kill all of the references to pyopenssl (other than historical ones) on the wiki now
19:50:00 <dopp> ok
19:50:09 <desai> lxml => libxml2
19:50:11 <desai> but other than that, no
19:50:33 <dopp> wait, so lxml and libxml2 cover the same base?
19:51:44 <desai> lxml uses libxml2
19:51:50 <dopp> ah, ok
19:51:55 <desai> it is actually python bindings for libxml2
19:52:08 <desai> since the libxml2 python bindings (that ship with libxml2) are pretty non-pythonic
19:52:31 <dopp> so the client needs libxml2 if it is using lxml instead of elementtree
20:16:00 <desai> yeah
20:47:33 <kisielk_mbp> desai: so for RPMng, what kind of repository do you need to host the packages?
20:48:26 <desai> anything rpm supports natively is fine, so nfs or http both work reasonably
20:49:20 <kisielk_mbp> will a plain HTTP server work? or does it need some special package listing or something on it?
20:50:21 <desai> plain http server is fine
20:50:30 <desai> urls get embedded in the config
20:54:04 <kisielk_mbp> cool
23:32:49 <kisielk_mbp> okay, so once I create a list of packages for Base
23:32:57 <kisielk_mbp> is there some way to generate all the Package entries ?
23:47:29 <Teknix> desai: not as part of the base installer, no
23:47:47 <Teknix> why?
23:54:55 <solj> kisielk_mbp: yes
23:55:10 <solj> using the pkgmgr_gen.py script
23:55:17 <solj> (for RPMs)