Coda File System

Re: need some help with replication

From: Deives Michellis <dmichell_at_grupogeo.com.br>
Date: Tue, 15 Jul 2003 14:45:10 -0300
First of all, I apoligize to you all, for I have forgotten to mention that I was using 5.3.20...

:)

When I said "n00b", I didnt mean newbie to linux... Actually, I'm a sysadmin and a C developer as well in my company.

I have a working openMosix cluster and was looking for a replicating filesystem to install cluster-wide, so it would be quite easy to create a really good fail-over system with high processing capacity... Perhaps you guys could tell me if coda is a good choice for that purpose, of if there is something else I should use (and stop annoying you :D)

thank you very much for your help and attention.



Deives

---

General System Failure:
"(A)bort, (R)etry, (G)et out the .45?"
"(A)bort, (R)etry, (I)gnore, (K)ick system"
"(A)bort, (R)etry, (I)nfluence with a large hammer."
"(A)bort, (R)etry, (P)retend it didn't happen."

On Tue, Jul 15, 2003 at 10:30:16AM -0400, Jan Harkes wrote:
> On Tue, Jul 15, 2003 at 09:59:06AM -0300, Deives Michellis wrote:
> > i need some help replicating coda fs...
> > 
> > every time I change those files, my "slave" coda servers die with
> > exitcode 77 and there is this message on the logs:
> 
> Which files? I'm guessing whenever your change /vice/db/VSGDB...
> 
> > oot_at_Trinity:/vice/srv# cat SrvErr
> > Assertion failed: 0, file "vsg.cc", line 222
> > EXITING! Bye!
> 
> Had a hard time finding this in my source tree, as the whole file got
> removed somewhere between 5.3.20 and 6.0.x. Good thing there is CVS...
> 
> This specific assertion triggers when you are trying to add a second
> volume storage group that has the same identifier as a group that was
> added previously.
> 
> > Here is my VSGDB:
> > 
> > root_at_Ragnarok:/vice/db# cat VSGDB 
> > E0000100 Ragnarok
> > E0000100 Trinity
> > E0000100 Ragnarok Trinity
> 
> Change this to,
> 
>   E0000100 Ragnarok
>   E0000101 Trinity
>   E0000102 Ragnarok Trinity
> 
> > I've tried diferente IDs (such as E0000101 Trinity and E0000102
> > Ragnarok Trinity), but the last ones didnt work. The volumes didnt
> > replicated.
> 
> You specify the replication group when you create the volumes and this
> creates all the underlying replicas and other administrivia.
> 
> Because you probably already have a couple volumes created with VSG
> E0000100 that are singly replicated on Ragnarok, you really can't just
> change which servers are a member of that group because the new server
> is missing the underlying replica and other information.
> 
> Use different VSGid's for different replication groups and specifically 
> create the volumes for a group however you want them replicated. With
> the above VSGDB the following command,
> 
>  createvol_rep coda:root_rep E0000102
> 
> Will create a new volume that has replicas on both ragnarok and trinity.
> Now if you want to use that volume as the volume that is mounted on
> /coda, you have to put that name in /vice/db/ROOTVOLUME and reinit your
> clients so that they correctly pick up the new root volume.
> 
> > Has anybody made a replication-howto-for-n00bs, or could send me you
> > config files as an example? I would really appreciate that, as the
> > documentation provided on the website is for advanced users, not
> > newbies.
> 
> I guess a lot of Coda isn't ready for newbies. There are too many
> configuration files to deal with and strange dependencies that will
> cause things to fail with incomprehensible errors (like the 'assert in
> vsg.cc:222' that you hit). But things are slowly getting better. The 6.0
> servers have no internal dependency on the VSGDB information, and at
> some point we will be able to say,
> 
>     createvol coda:root server1 server2 ...
> 
> And possibly,
> 
>     addreplica coda:root server3
>     removereplica coda:root server2
> 
> Jan
> 
> 
Received on 2003-07-15 13:46:32