[OpenAFS] Re: File creation delays

[Andrew Deason]

On Wed, 17 Mar 2010 14:43:30 -0400
"John W. Sopko Jr." <sopko@cs.unc.edu> wrote:

> Here is the strace command and the output of the FileLog trace, thanks
> for your help. I did a "rmdir try" and the strace command complained
> "No such device", if I "fs flushvolume" on the client it shows the
> directory got removed else it shows it still there.
> 
> Output from the "strace -tt rmdir try" command on the client host IP
> 152.2.140.200:
> 
> 14:25:39.851544 rmdir("try")            = -1 ENODEV (No such device)

I'm not sure why you get an ENODEV, but for the moment I'm ignoring it
to look at the delay problem instead... if anyone else has an intuition
on that, feel free.

> Here is the file server log "kill TSTP" 4 times during the same time,
> the client IP is 152.2.140.200

Trimming this down to the relevant thread...

> Wed Mar 17 14:25:39 2010 [12] SAFS_RemoveDir    try,  Did =
> 536884167.59.4927, Host 152.2.140.200:7001, Id 3903
> Wed Mar 17 14:25:39 2010 [12] BCB: BreakCallBack(all but
> 152.2.140.200:7001, (536884167,59,4927))
> Wed Mar 17 14:25:47 2010 [12] Starting multibreakcall back on all addr
> for host 152.2.140.115
> Wed Mar 17 14:25:54 2010 [12] BCB: Failed on file 536884167.59.4927,
> Host 152.2.140.115:7001 is down
> Wed Mar 17 14:25:54 2010 [12] SAFS_RemoveDir    returns 0

So, 152.2.140.115 looks like it's having trouble receiving callback
breaks. Do you know what that host is? Does it perchance have some kind
of firewall or anything that could prevent it from receiving incoming
UDP packets on port 7001?

-- 
Andrew Deason
adeason@sinenomine.net