---------------------- Forwarded by Yucel Erbilgic-YM/PGI on 11/14/2000 04:04 PM
---------------------------
From: Yucel Erbilgic-YM on 11/14/2000 03:55 PM
To: Ravi Channavajhala <ravi.channavajhala AT csfb DOT com>
cc:
Subject: Re: Re: [Veritas-bu] network errors (Document link: Yucel
Erbilgic-YM)
Hi Ravi,
I think your problem was due to the way unix communicates back to the client.
Your client can talk to MM on one interface but MM then depending on the load
can choose to answer back from any of its interfaces. If the MM talks to the
client on an interface that is not in client's bp.conf the client does not
answer, causing a time out.
2 solutions that I can think of at the moment could be available:
On the MM use required_interface command, for your backup backbone (but this
option is will not be usefull if you added this interface to load ballance
backups on both interface rather than seperating the production interface from
the backup, given netbackup will only utilze the interface specified I believe)
Or Your solution, add all the interfaces of te MM on the clients bp.conf. Then
you can possibly use the required_interface or /etc/hosts file on the clients
side to load ballance your 2 network interfaces.
I hope this sounds about right.
All have a nice week.
Regards,
Yucel M. Erbilgic
Internet Mail Message
Received from host: dns.eng.auburn.edu
[131.204.10.13]
Envelope Sender: veritas-bu-admin AT mailman.eng.auburn DOT edu
From: Ravi Channavajhala <ravi.channavajhala AT csfb DOT com> on 11/14/2000
02:40 PM
To: veritas-bu AT mailman.eng.auburn DOT edu
cc: (bcc: Yucel Erbilgic-YM/PGI)
Subject: Re: [Veritas-bu] network errors
For those who are interested, here is the real problem.
It just happened that I defined a new interface bkup02-e2
and plumbed it and brought it on line my media server when
we are having problems.
When I added the new interface bkup02-e2, I didnt update the
client's bp.conf. I was under the impression that clients
will use only the ones defined in their local bp.conf. However,
I defined the newly defined interface in master's bp.conf.
When a client backup request came, the master tried to send
them on bkup02-e2, causing the simple authentication based
on 'hostnames' failed and as a result, network errors - 54.
When I ran a traceroute on a client from media server, they
are going via bkup02-e2. So, I defined the bkup02-e2 in all
the failing client's bp.conf and all the problems disappeared.
Although my problem is solved, I would like to hear from someone,
why client's bp.conf entries are strictly not followed and how
does master make a determination to override the client's bp.conf
entries?
-ravi
On Fri, 3 Nov 2000, Ravi Channavajhala wrote:
ravi>Hi All,
ravi>
ravi>I am seeing a bunch of 41/54/57/59 errors from some of my NT clients
ravi>backing up to a Solaris Netbackup (3.2 GA) server. Often times, I
ravi>see the backup start and while in the middle of backup it times out
ravi>with error 54. I have upped teh CLIENT_CONNECT_TIMEOUT and
ravi>CLIENT_READ_TIMEOUT values in bp.conf to 1800 and extensively
ravi>trouble shooted the network.
ravi>
ravi>My setup is one media server, with several NICs, and each NIC
ravi>was defined in the bp.conf eg.,
ravi>
ravi>SERVER = bkup02
ravi>SERVER = bkup02-e0
ravi>SERVER = bkup02-f0
ravi>
ravi>Enabled verbose (99) for logging. Looking thru client's bpcd
ravi>shows that it connected to bkup02-f0, which is an interface
ravi>on the machine bkup02 and tries connecting to privilged ports
ravi>in the range of 600-900 and gives up after 20 tries with
ravi>
ravi>09:22:11 [258] <2> bpcd main: connection refused for port 882, try again
ravi>09:22:14 [258] <2> bpcd main: connection refused for port 883, try again
ravi>and so on...it tries a range of ports and gives up. Before this stage,
ravi>it acutually tells me that
ravi>
ravi>09:04:28 [258] <2> bpcd main: new fork cmd = <whatever>
ravi> <2> bpcd main: got socket for input 224
ravi> ....
ravi> <2> bpcd main: got socket for output 204, lport=883 and so
ravi>on..
ravi>
ravi>I upped the TCP timeout paramaters on the NT clients in the registry
ravi>and so on. Ran snoop and etc., without any luck. What is interesting
ravi>is sometimes the backup for the same client succeeds, and some times it
ravi>just plain and simple errors out with one of 41/54/57/59.
ravi>
ravi>Can someone give pointers to take the debugging and diagnosing this
ravi>to next level, beyond running 'bpclntcmd' from the client, verifying
ravi>the host name resolution, bp.conf etc., Thanks.
ravi>
ravi>-ravi
_______________________________________________
Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
|
