==> Regarding Re: [autofs] [RFC] rpc_ping and looong timeouts; Ian Kent <[EMAIL 
PROTECTED]> adds:

raven> On Thu, 3 Jun 2004, Jeff Moyer wrote:
>> Hello, Ian, list,
>> 
>> You may recall my asking for the rpc_ping function to take TCP only
>> servers into account.  Well, be careful what you ask for!  Now, if a
>> server is down in a replicated server configuration, it takes quite a
>> while to timeout due to the TCP connect.  The mount will eventually
>> succeed, but I've seen this take up to 10 minutes.

raven> I'm not sure that I see that same behaviour in my test environment.

raven> I'll check and get back.  It's probably best if we understand this a
raven> little better before jumping in.

I've strace'd the automount process, and a lot of time is spent in the
clnt_create, for both UDP and the TCP.  I implemented a non-blocking
connect, and it took my ls time down from 8m21s to 2m4s.  I'm still looking
into why the UDP stuff isn't timing out sooner.

>> Perhaps we should first test to see if a host is reachable at all before
>> we begin our rpc pings?

raven> Sigh!

raven> Will the fall out from the replicated server implementation it never
raven> end.

Despair not!  I'm sure we're on the home stretch, here!  ;)  I'll post a
patch for review when it's baked.

-Jeff

_______________________________________________
autofs mailing list
[EMAIL PROTECTED]
http://linux.kernel.org/mailman/listinfo/autofs

Reply via email to