[Gluster-users] One node goes offline, the other node can't see the replicated volume anymore

Brian Candler b.candler at pobox.com
Thu Jul 11 20:04:23 UTC 2013


On 11/07/2013 20:01, Greg Scott wrote:
>
> Well…OK. The “on my own” comment came from me after a long time and a 
> lot of work trying to figure this out.  I just went back and checked – 
> I posted the original question on 7/8 at 8:18 PM. I asked a follow-up 
> question 7/9 at 12:27 PM, roughly 16 hours later.  And then the “On my 
> own” post was on 7/10 at 12:27 AM, or around 28 hours after my 
> original question.   I was feeling kind of lonely and, well, on my own 
> at the time.  All times are USA Central time.  I do work weird hours.
>
I believe the trolling comment was aimed at me, although what I actually 
said was less polite than "you're on your own" :-)

I take the point that there is an active and helpful gluster support 
community, but IMO that is papering over the underlying cracks. Any data 
storage system must be built to hand the error cases first and foremost 
- performance and features are secondary. What's the point of having any 
sort of redundant system which happily copies your data when everything 
is working fine, but then gets itself into a twist when there's a problem?

Without the documentation of the algorithms and states, there's no way 
to dig yourself out manually either. Gluster used to have a half-decent 
wiki in the 2.x days, but that was orphaned, and now basically the 
technical internal documentations consists of a couple of blog postings.

You have the source code of course, but for various pluggability and 
performance reasons, gluster's source is hard to get one's head around 
(I couldn't anyway). And even then, it shouldn't be the norm to have to 
read source code when things break in a production environment. The 
procedures should be documented - where to look, what you may see in 
different scenarios, what actions you should take in those scenarios.

I apologise if my remarks came across as facetious, but to be honest, I 
wish someone had pointed out the issues before I had started on the 
journey into gluster and back out again. Since making that posting, I've 
had a couple of private replies in support of what I said. Maybe it's 
considered impolite to say anything negative about a project on that 
project's mailing list.

I remain on the list, mostly lurking, to see if one day things improve. 
However if someone will set up a gluster-ex-users mailing list, I'll 
join that :-)

Regards,

Brian.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130711/47a74ab5/attachment.html>


More information about the Gluster-users mailing list