16:23:00 <Eco_> #startmeeting
16:23:00 <zodbot> Meeting started Thu Nov  1 16:23:00 2012 UTC.  The chair is Eco_. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:23:00 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
16:23:21 <Eco_> is there anyone besides you and I semiosis?
16:23:28 <semiosis> JoeJulian: ping
16:23:35 <semiosis> jdarcy: ping
16:23:38 <semiosis> we'll see
16:23:53 <Eco_> a2, jdarcy
16:23:55 <Eco_> ?
16:26:18 <Eco_> we'll give it a few minutes, not much on the agenda for this week either way
16:26:51 <Eco_> morning UnixDev
16:27:12 <UnixDev> morning :)
16:27:35 <Eco_> we are just waiting a few more minutes to see if anyone else joins
16:29:54 <Eco_> UnixDev, I have seen and talked to you in #gluster recently, have you attended any of the infra meetings?
16:30:09 <UnixDev> nope, this will be my first
16:30:22 <Eco_> semiosis, i think this is it for this week
16:30:51 <Eco_> as a plus, we have increased our membership from last time johnmark was out by 50%....welcome again, UnixDev ;)
16:31:13 <UnixDev> lol, nice, thank you
16:32:17 <Eco_> lol  no problem....its actually serendipitous that you joined since you (indirectly) are involved in one of the topics today
16:32:42 <Eco_> which leads us to our first topic...
16:32:46 <Eco_> case studies!
16:33:35 <Eco_> semiosis, were you able to check out any questionnaire software that would be suitable for our needs?
16:34:02 <Eco_> UnixDev, from last week, we were talking about case studies
16:34:12 <semiosis> back
16:34:19 <Eco_> you may be a great candidate for this
16:34:22 <Eco_> welcome back
16:34:41 <Eco_> UnixDev, since you had a lot of questions about VMware and Gluster together
16:35:14 <Eco_> since there are plenty of people who would be interested, we would love your feedback on what you have done to be successful with it
16:36:12 <Eco_> and of course would be happy to give what pointers we can to help you be successful with it
16:36:14 <UnixDev> great, I have lots of this!
16:36:34 <UnixDev> give me a moment, I'm having an unrelated issue here
16:36:47 <Eco_> so, semiosis is taking the lead on case studies for community users....sure, no problem
16:36:48 <semiosis> Eco_: not yet
16:36:57 <semiosis> still working on it
16:37:13 <Eco_> semiosis, no worries, thanks for the update
16:38:32 <Eco_> can you give UnixDev some background on what we are looking for with the case studies when he gets back?  this seems to me to be an excellent fit
16:38:59 <Eco_> we can use him as our guinea pi....errr...valued contributor!  ;-)
16:42:41 <semiosis> examples of successful deployments of glusterfs in various use-cases, for example, dynamic web apps, streaming media, vm disk images, backup/archival, large scratch fs, nfs with vip, hadoop, ufo
16:43:55 <Eco_> well, i think this is a perfect fit then of course
16:44:27 <Eco_> vm images in ESX 5 (and whatever they are calling VI these days)
16:47:41 <UnixDev> so yes, we are running esx 5.1
16:48:09 <UnixDev> and I'm using gluster to store vm images (right now we are using nfs) but vmware likes iscsi a bit more
16:48:54 <UnixDev> i tried creating iscsi targets from files on a gluster vol, but that didn't go so well.. maybe it was the 500 gig files that were used? perhaps gluster is more for smaller files?
16:49:19 <semiosis> well... did you try it with smaller files?
16:49:25 <Eco_> UnixDev, there may have been other issues as well, but typically gluster works better for large files
16:49:33 <UnixDev> yes, right now with nfs, the disks are each files
16:49:39 <semiosis> we can discuss that & your experience w/ iscsi more in #gluster
16:49:48 <UnixDev> got it, so regarding nfs
16:50:21 <UnixDev> the biggest issue for me is HA. vmware does not have multipathing for nfs
16:50:51 <UnixDev> another issue is the split log, thats confusing.. and just healing split-brain in general
16:51:09 <Eco_> UnixDev, have you written up the experiences you have had so far with it?
16:51:15 <UnixDev> it also seems that when using heartbeat in a failover scenario, gluster needs to be restarted before responding in the newly added ip
16:51:38 <UnixDev> Eco_: I've opened some bugs, one for split brain logs, another for some striping enhancement
16:52:17 <UnixDev> Eco_: i have some notes, but not really a formal writeup … I wish it had gone better to have done the writeup personally
16:53:42 <Eco_> UnixDev, since part of what we want to do with the case studies is to start socializing failure as a _positive_ thing, it would be good to have your thoughts on the issue as it stands now
16:54:19 <Eco_> by doing that, you help others avoid the same mistakes, but it also becomes a great story if we are able to overcome the hurdles
16:54:33 <UnixDev> sure, well I like gluster, there are many great things and I think the potential is there
16:55:05 <Eco_> we are running short on time, semiosis did you have more that you wanted to say about the case study for now?  do you agree this is a good first test case?
16:55:14 <UnixDev> I can write up some info on failed paths I took, but maybe even those I can explore with you guys as there could be some tweaks that would make then work correctly
16:55:22 <semiosis> sure
16:56:44 <semiosis> while i agree that documenting failures is helpful for improving glusterfs, that's a long-term goal & not really the point of the case studies imho
16:57:13 <semiosis> i think the point of the case studies is to help people get going today with a plan to run glusterfs in production
16:57:41 <semiosis> case studies that end in or focus too much on failed strategies won't help people get up & running
16:57:55 <UnixDev> it would be great for people wanting to use vmware… there are little docs about gluster with that
16:58:12 <UnixDev> i had to start building form source because packages and a bug with vmware where vm's would not start
16:58:14 <semiosis> agreed
16:58:54 <Eco_> semiosis, i am for whatever way you want to go on it
16:59:04 <JoeJulian> +1
16:59:10 <UnixDev> it would also be a good idea to point out the master branch is not stable? I think someone in the channel mentioned this.. maybe some better info about source branches on website
16:59:26 <semiosis> if a plan to run glusterfs in production begins with "first fix or change all this stuff in glusterfs" that's not a good plan, i would not approve that if someone on my team proposed it for our prod infra
16:59:36 <Eco_> can you work with UnixDev as the first use case for this, whatever end result you want to see is really what i am after
16:59:37 <semiosis> :)
16:59:42 <Eco_> JoeJulian, welcome
17:00:11 <Eco_> ok, we are out of time, there were just a few more items i wanted to mention
17:00:37 <Eco_> johnmark and I will both be out tomorrow so there will be no community hours this week
17:00:45 <semiosis> UnixDev: lets chat more about your use case in #gluster, will you hang around for a few days so we can work on this?
17:01:06 <Eco_> most likely not for next week either, which leads me to my last topic
17:01:11 <UnixDev> yeah sure, I think i may have also found some sparse file bugs
17:01:29 <Eco_> JoeJulian, glad you are here since i think this is something you would like to do
17:02:34 <Eco_> we wanted to get volunteers for the community hours to present on topics, is there anything you want to present to that end?
17:02:43 <Eco_> same goes for you as well semiosis
17:03:01 <JoeJulian> Nothing for tomorrow, but I can come up with something for next week.
17:03:54 <JoeJulian> If I had a month to plan, maybe, but your steel, his strength and my brain against 60 men, impossible!
17:04:05 <JoeJulian> ... if I only had a wheelbarrow....
17:04:21 <jdarcy> ...and a holocaust cloak...
17:05:01 <Eco_> JoeJulian, jdarcy, you both made my morning
17:05:08 <Eco_> doh
17:05:10 <Eco_> should have said
17:05:17 <Eco_> stay on topic!  i mean it!
17:05:23 <Eco_> anybody want a peanut?
17:05:26 <JoeJulian> anybody want a peanut?
17:05:39 <Eco_> JoeJulian, the time frame is for whenever you want
17:06:21 <Eco_> jdarcy, you would also be tops on the guest list since, life EF Hutton, when jdarcy speaks, people listen!
17:06:59 <jdarcy> Yes, some time I need to get Hangouts working so I can do that.  ;)
17:07:17 <Eco_> jdarcy, johnmark can add you to the existing one
17:07:53 <Eco_> if its a matter of actually working on a specific device, i am not likely to be much help but i can tell you the ways *I* have failed doing it  ;)
17:08:39 <Eco_> that's it for this week from my side, is there anything else anyone would like to mention / discuss / parlay?
17:11:58 <Eco_> ok, looks like not
17:12:07 <Eco_> with that, i bid you all adieu
17:12:10 <Eco_> #endmeeting