==> Regarding RE: [autofs] automounter failing with NetApp snapmirror; [EMAIL 
PROTECTED] adds:

raven> On Mon, 19 Jul 2004, Jeff Moyer wrote:
>> ==> Regarding RE: [autofs] automounter failing with NetApp snapmirror;
>> [EMAIL PROTECTED] adds:
>> 
raven> On Mon, 19 Jul 2004, Jeff Moyer wrote:
>> >> ==> Regarding RE: [autofs] automounter failing with NetApp
>> snapmirror; >> [EMAIL PROTECTED] adds:
>> >> 
raven> On Sat, 17 Jul 2004, Lever, Charles wrote:
>> >> >> > > hi all-
>> >> >> > > 
>> >> >> > > i'm still learning about the Linux automounter, but we have >
>> an >> >> issue > > that seems to come up a lot for our customers using
>> Linux, >> >> it's > > automounter, and NetApp filers doing volume
>> snapmirror (a > >> >> file system > > replication mechanism).
>> >> >> > > 
>> >> >> > > at the end of a replication event, the destination file system
>> >> goes >> > > offline for 3-4 seconds while the new version replaces
>> it.  >> > the >> filer > > used to return EACCES during this period, but
>> we've >> changed it >> to > > simply ignore all RPC requests for that
>> file system >> until the new >> > > version is online, as this seems to
>> be more >> client-friendly > >> behavior.
>> >> >> > > 
>> >> >> > > so now, during that time period, an automounter request to >
>> >> mount >> that > > file system will fail with "RPC: timed out" rather
>> >> than getting >> > > EACCES. is there any way we can tell the
>> automounter >> to > wait >> longer or > > retransmit another few times?
>> >> >> > > 
>> >> 
>> >> I believe that error is produced by the mount command, itself.
>> Poking >> through the code for mount, it doesn't look like this timeout
>> is >> configurable.  Autofs 4 actually checks to see if a host is alive
>> before >> attempting a mount, so at least things should work there.  I'm
>> not sure >> what the best approach is to solve this problem on 3.1.7.
>> >> 
>> 
raven> But due to complaints, the check is only done against entries that
raven> use replicated server syntax. Also, the latest code is not yet
raven> included in a release.
>> Ahh, forgot about that.  BTW, the code is present in FC3 at least, and
>> soon RHEL 3.
>> 
>> >> >> > >> > autofs version?  > kernel version?  > patched/not patched?
>> > >> if >> patched which revision?
>> >> >> 
>> >> >> the specific cases we're seeing right now are with RHEL AS 2.1, >>
>> updates >> 2 and 3.  jeff moyer can provide you with the exact patch set
>> >> used in >> that distribution.
>> >> 
raven> I expect that's autofs version 3 and the stock version 3 kernel
raven> module.
>> >> AS2.1 Update 3 has autofs version 3.1.7-21.  The kernel module likely
>> >> matches the code in vanilla 2.4.9, but I don't believe this code to
>> be >> suspect.
>> >> 
raven> It's worth pointing out that the v3 code has not changed for a long
raven> time so perhaps the source of this is somewhere else.
>> I'm not sure I follow you, here, Ian.  This is new behaviour from the
>> server that needs handling.  It's precisely because the code hasn't
>> changed that we have a problem, no?  ;)

raven> I guess, but don't the RH AS releases remain functionaly the same
raven> between releases to reduce this type of issue. In which case my
raven> question is which update caused the problem? If we can zero in on it
raven> maybe we can work around it easily.

Yes, we try to remain functionally the same between releases.  However, in
this case the server is a NetApp.  It was a change in NetApp behaviour
that got us into this situation, not RHEL software.

-Jeff

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

Reply via email to