Re: [Veritas-bu] Master server can't see media servers
2007-07-06 13:28:29
Oh my.. my addresses was listed.. Forgot to remove it. Btw, if you still not sure where's the problem, and you said that the /usr/openv was mounted from another disk via FC connection, and your OS solaris 10, I did had almost same environment..
could be the solaris 10 tcp fusion issue. you should turn it off. until SUN release patches for this issue. my case was the /usr/openv/netbackup mounted on FC to another disk. And everytime we do the catalog backup, it always failed.
Try to add this on your /etc/system file : "set ip:do_tcp_fusion = 0" (without quotes) then do the reboot -- -r after that, my catalog backup was run properly.. :D hopefully works on you Michael..
@ Ed: Gee.. I dont know about that new features on 6.0 :) Do you know how the maximum media server we can have on NBU 6 Ed? Thanks for da info.. :D
Regards, mTz
On 7/6/07, Ed Wilts <ewilts AT ewilts DOT org> wrote:
50+ media servers should not be that risky with 6.0 – one of the
driving reasons behind the 6.0 release was for scalability, especially with a
large number of media servers.
Are the pbx processes all running properly? If they weren't,
then I could picture the symptoms you're seeing with ssh/ping working, but
NetBackup not working.
…/Ed
Yeah, I know having 50+ media servers is risky....but there really
is no other way to properly do it.
I'm beginning to think it isn't netbackup. My /usr/openv
directory is mounted via a direct connect fibre on a Sun 9990 raid. I'm
running veritas file system (vxfs) 5.0. I'm also doing Veritas Volume
Replicator (VVR) to another Sun 9990 raid at a DRS site. Both sites are
connected via a large, private pipe...so bandwidth between the two sites isn't
an issue. But I'm seeing this in my /var/adm/messages file:
vxio V-5-0-0 disconnecting rlink rlk_<hostname>_bkprvg due to
exessive retries.
Also...when some stuff hangs, and I go into /usr/openv (or any
directory under that) my terminal hangs. I can ssh back into the
box, and I'm okay until I navigate into /usr/openv. I've done a full
fsck of the vxfs file system, it wasn't clean the first time...but has come
back clean since.
So I may be having vxfs issues. :-/ Yippie....
--
Mike Sponsler
Michael.Sponsler AT ngc DOT com
Northrop Grumman Information Technology
From: Martin Ruslan [mailto:
mit.martin AT gmail DOT com]
Sent: Friday, July 06, 2007 1:49 AM
To: veritas-bu AT mailman.eng.auburn DOT edu
Cc: Sponsler, Michael
Subject: Re: [Veritas-bu] Master server can't see media servers
Yeap..
it's odd.. :)
well.. as far as I know, it's too risky if you have that much of media server.
Because they alway communicate each other, and when even one media server
couldn't talk, the process will be hung.
Are you already check this:
- on /usr/openv/netbackup/bp.conf, did all the media server
registered there, with the "SERVER = media_server_name" (without
quotes) ?
yes
for all of the media server too?
check on: /usr/openv/netbackup/bin/bpps -a for the hung processes.
try to kill the hung process with ./kill -9 <Hunged PID process>
Then you'll know which media server had the problems.
Regards,
mTz
On 7/6/07, Sponsler, Michael
<Michael.Sponsler AT ngc DOT com>
wrote:
- Did all of your
media server registerd on your /etc/hosts file at the master server?
yes. The environment had been working for several months. No recent
patch updates or changes (that I'm aware of). The master server lost
"netbackup communication" with the media servers upon restarting the
netbackup daemons.
- did the master server name and ip address listed on
/etc/hosts at all the media server?
yes. I can ping and ssh to all media servers and vice versa
- on /usr/openv/netbackup/bp.conf, did all the media server
registered there, with the "SERVER = media_server_name" (without
quotes) ?
yes
- check on the media server:
"./usr/openv/volmgr/bin/vmglob get_gdbhost"
It gives me the master server's
hostname
It's odd, huh?
--
Mike Sponsler
Michael.Sponsler AT ngc DOT com
Northrop Grumman Information Technology
From: Martin Ruslan [mailto:
mit.martin AT gmail DOT com ]
Sent: Friday, July 06, 2007 1:36 AM
To: veritas-bu AT mailman.eng.auburn DOT edu
Cc: Sponsler, Michael
Subject: Re: [Veritas-bu] Master server can't see media servers
- Did all of your
media server registerd on your /etc/hosts file at the master server?
- did the master server name and ip address listed on /etc/hosts
at all the media server?
- on /usr/openv/netbackup/bp.conf, did all the media server
registered there, with the "SERVER = media_server_name" (without
quotes) ?
- check on the media server:
"./usr/openv/volmgr/bin/vmglob get_gdbhost"
is the result was the master server? if not, run:
"./usr/openv/volmgr/bin/vmglob set_gdbhost master_server_name"
Check it, and give us the result.. :)
Regards,
mTz
On 7/6/07, Sponsler, Michael
<Michael.Sponsler AT ngc DOT com
> wrote:
Netbackup 6.0 MP4, solaris 10 master server; Netbackup 6.0
MP4, solaris 8 san media servers
Roughly 55 media servers.
After I rebooted the netbackup daemons, I had the following issue:
Master server can ping and ssh to all media servers, but Master
server cannot communicate to media servers via netbackup. vmoprcmd
command hangs. Any jobs for media servers that start up come back with
"Media server is not active". Netbackup is running on master
server and all media servers. Tried rebooting the master server with same
outcome. There is no firewall between the master server and any media
servers.
Something obviously changed before I restarted the Netbackup
daemons....anyone have any ideas?
Northrop Grumman Information Technology
_______________________________________________ Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|
|
|