[OpenAFS] Disk cache not updating
Brian Huntley
bhuntley@clarkson.edu
Mon, 10 Jan 2005 16:12:30 -0500
This is a multipart message in MIME format.
--=_alternative 00745F3A85256F85_=
Content-Type: text/plain; charset="US-ASCII"
Thanks...
I'm at 1.2.11 ( should move to 2.13- I thought I was already at current -
sorry). When I see the problem, a flushvol does produce a callback, and
it continues to work for some period of time, and then begins failing
again. When it is failing, the volume doesn't show up in a cmdebug at
all, and all of the entries that are there have positive expirations.
In terms of blocking messages, I have ports 7000-7009 and 1024-2048 open
on both server and client. There's only one hop between the two on the
network, and there isn't any filtering or ACL's going on at that router.
Best,
Brian
--
Brian T. Huntley, Manager of Network Operations
Office of Information Technology
Clarkson University
bhuntley@clarkson.edu -- 315.268.6723
"UNIX *is* user friendly. It's just selective about who its friends are."
Jeffrey Altman <jaltman@columbia.edu>
Sent by: openafs-info-admin@openafs.org
01/10/2005 11:36 AM
To
Brian Huntley <bhuntley@clarkson.edu>
cc
openafs-info@openafs.org
Subject
Re: [OpenAFS] Disk cache not updating
Brian:
Useful information would be the OpenAFS version number you have
installed. I have no idea what is the latest OpenAFS rpm available
translates to.
Does it appear from the output of "cmdebug <host> -long" that you
have callbacks for volume in question? cmdebug lists cache entries
with an ID of the form: <cellID>.<volID>.<vnode>.<unique> Using
volume ID you can search the output for entries from the specified
volume and look to see if there is a callback address and a non-zero
expiration time.
Do you have a callback after the "fs flushvol" command is issued?
If you have a callback after the "fs flushvol" and it does not get
renewed at expiration, then there is a bug somewhere. Perhaps you
are correct in that it is triggered by load. If so it might be
caused by reference count overflow. Do any of the reference counts
in the cmdebug output display as negative?
If you do have callbacks and do not see the changes, the second
most likely cause is that the messages from the afs file server
to the client are blocked somewhere.
Jeffrey Altman
Brian Huntley wrote:
>
> Hello everyone...
>
> I can't seem to find anything on this particular type of cache problem,
> so hopefully someone can point me in the right direction.
>
> We recently put into production a webserver whose document root is in
> AFS. During the development of this system, we never had an issue with
> the AFS client keeping it's cache current with what was on the server.
> Since it's been put into production, however, we are finding that we
> frequently must use fs flushvol to keep the cache up to date with the
> AFS server's version of volumes. It seems that the volumes most heavily
> hit by apache are the ones that most often need a flush to stay current.
>
> I can only hypothesize that this is therefore a transaction/load related
> issue. The webserver itself is virtually idle, but serves tens of
> thousands of requests per day.
>
> We're running RedHat Enterprise 3, Update 3 and the latest OpenAFS rpm
> available for it.
>
> Any thoughts?
>
> TIA!
> -b
> --
> Brian T. Huntley, Manager of Network Operations
> Office of Information Technology
> Clarkson University
> bhuntley@clarkson.edu -- 315.268.6723
> "UNIX *is* user friendly. It's just selective about who its friends
are."
_______________________________________________
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info
--=_alternative 00745F3A85256F85_=
Content-Type: text/html; charset="US-ASCII"
<br><font size=2 face="sans-serif">Thanks...</font>
<br>
<br><font size=2 face="sans-serif">I'm at 1.2.11 ( should move to 2.13-
I thought I was already at current - sorry). When I see the problem,
a flushvol does produce a callback, and it continues to work for some period
of time, and then begins failing again. When it is failing, the volume
doesn't show up in a cmdebug at all, and all of the entries that are there
have positive expirations.</font>
<br>
<br><font size=2 face="sans-serif">In terms of blocking messages, I have
ports 7000-7009 and 1024-2048 open on both server and client. There's
only one hop between the two on the network, and there isn't any filtering
or ACL's going on at that router.</font>
<br>
<br>
<br><font size=2 face="sans-serif">Best,</font>
<br><font size=2 face="sans-serif">Brian</font>
<br><font size=2 face="sans-serif">--<br>
Brian T. Huntley, Manager of Network Operations<br>
Office of Information Technology<br>
Clarkson University<br>
bhuntley@clarkson.edu -- 315.268.6723<br>
"UNIX *is* user friendly. It's just selective about who its friends
are."<br>
</font>
<br>
<br>
<br>
<table width=100%>
<tr valign=top>
<td width=40%><font size=1 face="sans-serif"><b>Jeffrey Altman <jaltman@columbia.edu></b>
</font>
<br><font size=1 face="sans-serif">Sent by: openafs-info-admin@openafs.org</font>
<p><font size=1 face="sans-serif">01/10/2005 11:36 AM</font>
<td width=59%>
<table width=100%>
<tr valign=top>
<td>
<div align=right><font size=1 face="sans-serif">To</font></div>
<td><font size=1 face="sans-serif">Brian Huntley <bhuntley@clarkson.edu></font>
<tr valign=top>
<td>
<div align=right><font size=1 face="sans-serif">cc</font></div>
<td><font size=1 face="sans-serif">openafs-info@openafs.org</font>
<tr valign=top>
<td>
<div align=right><font size=1 face="sans-serif">Subject</font></div>
<td><font size=1 face="sans-serif">Re: [OpenAFS] Disk cache not updating</font></table>
<br>
<table>
<tr valign=top>
<td>
<td></table>
<br></table>
<br>
<br>
<br><font size=2><tt>Brian:<br>
<br>
Useful information would be the OpenAFS version number you have <br>
installed. I have no idea what is the latest OpenAFS rpm available<br>
translates to.<br>
<br>
Does it appear from the output of "cmdebug <host> -long"
that you<br>
have callbacks for volume in question? cmdebug lists cache entries<br>
with an ID of the form: <cellID>.<volID>.<vnode>.<unique>
Using<br>
volume ID you can search the output for entries from the specified<br>
volume and look to see if there is a callback address and a non-zero<br>
expiration time.<br>
<br>
Do you have a callback after the "fs flushvol" command is issued?<br>
<br>
If you have a callback after the "fs flushvol" and it does not
get<br>
renewed at expiration, then there is a bug somewhere. Perhaps you<br>
are correct in that it is triggered by load. If so it might be<br>
caused by reference count overflow. Do any of the reference counts<br>
in the cmdebug output display as negative?<br>
<br>
If you do have callbacks and do not see the changes, the second<br>
most likely cause is that the messages from the afs file server<br>
to the client are blocked somewhere.<br>
<br>
Jeffrey Altman<br>
<br>
<br>
<br>
Brian Huntley wrote:<br>
<br>
> <br>
> Hello everyone...<br>
> <br>
> I can't seem to find anything on this particular type of cache problem,
<br>
> so hopefully someone can point me in the right direction.<br>
> <br>
> We recently put into production a webserver whose document root is
in <br>
> AFS. During the development of this system, we never had an
issue with <br>
> the AFS client keeping it's cache current with what was on the server.
<br>
> Since it's been put into production, however, we are finding
that we <br>
> frequently must use fs flushvol to keep the cache up to date with
the <br>
> AFS server's version of volumes. It seems that the volumes most
heavily <br>
> hit by apache are the ones that most often need a flush to stay current.<br>
> <br>
> I can only hypothesize that this is therefore a transaction/load related
<br>
> issue. The webserver itself is virtually idle, but serves tens
of <br>
> thousands of requests per day.<br>
> <br>
> We're running RedHat Enterprise 3, Update 3 and the latest OpenAFS
rpm <br>
> available for it.<br>
> <br>
> Any thoughts?<br>
> <br>
> TIA!<br>
> -b<br>
> --<br>
> Brian T. Huntley, Manager of Network Operations<br>
> Office of Information Technology<br>
> Clarkson University<br>
> bhuntley@clarkson.edu -- 315.268.6723<br>
> "UNIX *is* user friendly. It's just selective about who its friends
are."<br>
_______________________________________________<br>
OpenAFS-info mailing list<br>
OpenAFS-info@openafs.org<br>
https://lists.openafs.org/mailman/listinfo/openafs-info<br>
</tt></font>
<br>
--=_alternative 00745F3A85256F85_=--