expire_indirect errors in the syslog

expire_indirect errors in the syslog

[Ondrej Valousek]

Hi all,

I am seeing quite lot of following messages in the syslog:
Oct 21 02:17:05 ara automount[3181]: expire_indirect: fstat failed: Bad
file descriptor

On RHEL-5.2, last updates, autofs-5.0.1-0.rc2.88.
I have seen some discussion on the net (bug #448038) regarding this
error and what is common is, that my  NFS server is using the "nohide"
option.

Question - is this bug dangerous or not? Is there any resolution to
this? Recently I have experienced our logon server instability (Gnome
session frozen) due to a problem with gnome-vfs so I do not know if this
is related....
Thanks,

Ondrej

Re: expire_indirect errors in the syslog

[Ian Kent]

On Tue, 2008-10-28 at 08:46 +0100, Ondrej Valousek wrote:
> Hi all,
> 
> I am seeing quite lot of following messages in the syslog:
> Oct 21 02:17:05 ara automount[3181]: expire_indirect: fstat failed: Bad
> file descriptor
> 
> On RHEL-5.2, last updates, autofs-5.0.1-0.rc2.88.
> I have seen some discussion on the net (bug #448038) regarding this
> error and what is common is, that my  NFS server is using the "nohide"
> option.
> 
> Question - is this bug dangerous or not? Is there any resolution to
> this? Recently I have experienced our logon server instability (Gnome
> session frozen) due to a problem with gnome-vfs so I do not know if this
> is related....

Ummm .. it could be but should only be an issue when related to the
"nohide" handling.

Log a bug and get hold of rev 101 and see how things go.

Ian

Re: expire_indirect errors in the syslog

[Ondrej Valousek]

Hi Ian,

Thanks for the quick reply. 2 more questions:
1. Is the "nohide" bug exposed even when using NFSv3 (NFSv3 client
mounting the nohide nfs server export)
2. from the bug #448038 details I can see a patch for this has already
been accepted and will go into RHEL-5. When?

I have already filed a bug regarding the gnome-vfs, but I do not know
whether I should also mention the autofs errors.
I do not need NFSv4 right now - I just wanted to prepare my NFS servers
for v4 - so I introduced the nohide option.
For sake of safety, I have removed this option for now.

Thanks,
Ondrej

Ian Kent wrote:
> On Tue, 2008-10-28 at 08:46 +0100, Ondrej Valousek wrote:
>   
>> Hi all,
>>
>> I am seeing quite lot of following messages in the syslog:
>> Oct 21 02:17:05 ara automount[3181]: expire_indirect: fstat failed: Bad
>> file descriptor
>>
>> On RHEL-5.2, last updates, autofs-5.0.1-0.rc2.88.
>> I have seen some discussion on the net (bug #448038) regarding this
>> error and what is common is, that my  NFS server is using the "nohide"
>> option.
>>
>> Question - is this bug dangerous or not? Is there any resolution to
>> this? Recently I have experienced our logon server instability (Gnome
>> session frozen) due to a problem with gnome-vfs so I do not know if this
>> is related....
>>     
>
> Ummm .. it could be but should only be an issue when related to the
> "nohide" handling.
>
> Log a bug and get hold of rev 101 and see how things go.
>
> Ian
>
>
>   

Re: expire_indirect errors in the syslog

[Ian Kent]

On Tue, 2008-10-28 at 09:19 +0100, Ondrej Valousek wrote:
> Hi Ian,
> 
> Thanks for the quick reply. 2 more questions:
> 1. Is the "nohide" bug exposed even when using NFSv3 (NFSv3 client
> mounting the nohide nfs server export)

Yes. If your map is mounting based on exports from the server there is
no way to know, from the export information, if the NFS client will will
automatically mount the export (the "nohide" option) which leads to the
conflict.

> 2. from the bug #448038 details I can see a patch for this has already
> been accepted and will go into RHEL-5. When?

5.3.

> 
> I have already filed a bug regarding the gnome-vfs, but I do not know
> whether I should also mention the autofs errors.
> I do not need NFSv4 right now - I just wanted to prepare my NFS servers
> for v4 - so I introduced the nohide option.
> For sake of safety, I have removed this option for now.
> 
> Thanks,
> Ondrej
> 
> Ian Kent wrote:
> > On Tue, 2008-10-28 at 08:46 +0100, Ondrej Valousek wrote:
> >   
> >> Hi all,
> >>
> >> I am seeing quite lot of following messages in the syslog:
> >> Oct 21 02:17:05 ara automount[3181]: expire_indirect: fstat failed: Bad
> >> file descriptor
> >>
> >> On RHEL-5.2, last updates, autofs-5.0.1-0.rc2.88.
> >> I have seen some discussion on the net (bug #448038) regarding this
> >> error and what is common is, that my  NFS server is using the "nohide"
> >> option.
> >>
> >> Question - is this bug dangerous or not? Is there any resolution to
> >> this? Recently I have experienced our logon server instability (Gnome
> >> session frozen) due to a problem with gnome-vfs so I do not know if this
> >> is related....
> >>     
> >
> > Ummm .. it could be but should only be an issue when related to the
> > "nohide" handling.
> >
> > Log a bug and get hold of rev 101 and see how things go.
> >
> > Ian
> >
> >
> >   
> 

Re: expire_indirect errors in the syslog

[Ondrej Valousek]

Ok, I have removed all "nohide" options from my exports, but I am still
experiencing the expire_indirect errors.
I must also say that I am also using NetApp NAS and one share is nfs4
mounted.
But I do not believe it is NetApp/nfs4 related.

Question:
Is it possible that I receive this errors when:
1. say I have autofs key like "share nfsserver:/exports/share"
2. some processes start using the share
3. I delete the key above and delete the share from the NFS server
4. As the share no longer exists but some processes are still using it,
automounter can not unmount it, so I manually invoke umount -l to remove it.

Thanks,
Ondrej

Re: expire_indirect errors in the syslog

[Ian Kent]

On Thu, 2008-11-06 at 13:22 +0100, Ondrej Valousek wrote:
> Ok, I have removed all "nohide" options from my exports, but I am still
> experiencing the expire_indirect errors.
> I must also say that I am also using NetApp NAS and one share is nfs4
> mounted.
> But I do not believe it is NetApp/nfs4 related.
> 
> Question:
> Is it possible that I receive this errors when:
> 1. say I have autofs key like "share nfsserver:/exports/share"
> 2. some processes start using the share
> 3. I delete the key above and delete the share from the NFS server
> 4. As the share no longer exists but some processes are still using it,
> automounter can not unmount it, so I manually invoke umount -l to remove it.

This specific example should not result in those messages.
If you want further analysis of this issue then post a full debug log.

Ian

Re: expire_indirect errors in the syslog

[Ondrej Valousek]

Ian,

I guess I can not turn debugging on without restarting autofs, right?
This is a production login box with many users logged in. My experience
says restarting autofs would freeze the machine.
Thanks,

Ondrej
> This specific example should not result in those messages.
> If you want further analysis of this issue then post a full debug log.
>
> Ian
>
>
>   

Re: expire_indirect errors in the syslog

[Ian Kent]

On Fri, 2008-11-07 at 08:42 +0100, Ondrej Valousek wrote:
> Ian,
> 
> I guess I can not turn debugging on without restarting autofs, right?
> This is a production login box with many users logged in. My experience
> says restarting autofs would freeze the machine.

That can be a problem, I wouldn't.

But occasionally people report enabling debugging on heavily loaded
systems can cause them to behave badly. So, even though you can enable
debug logging on the fly (check the man page), I wouldn't do that
either.

But that also means that, if a problem has already happened, then
changing the server exports likely won't fix it either. But of course it
depends on the problem.

If your not having any specific problems, other than annoying log
messages, then leave the machine alone.

> Thanks,
> 
> Ondrej
> > This specific example should not result in those messages.
> > If you want further analysis of this issue then post a full debug log.
> >
> > Ian
> >
> >
> >   
>