Autofs 4.1.4_beta2 + NetApp Filer + missing dir

Autofs 4.1.4_beta2 + NetApp Filer + missing dir

[devnull]

Hello All,

This is an extension to the "unacceptable bug in autofs kernel module" 
sent by mmarion at qualcomm

http://linux.kernel.org/pipermail/autofs/2005-February/001889.html

Our situation

/usr/sbin/automount -V
Linux automount version 4.1.4_beta2
(4.1.2 didnt work either)

We are submitting jobs to machines via LSF, the data being stored on 
NetApp filers, running OnTap 6.5.2

We run a perl script from

/home/user/workarea called test.pl

test.pl basically does


1. setup a bunch of stuff for our runs, including versions etc.

2. rmtree "DirectoryNeeded" (to make sure we remove cached data from older
    runs)

3. bsub the job, wait for completion(this job has now created a new copy 
of "DirectoryNeeded")

4. Once the bsub completes, we then try to do an "opendir" on
    "DirectoryNeeded" and it doesnt work. Just says not found.


To make the script work I can

1. Take out Step 2(the rmtree part), remove the dir manually and run the 
script and it works great.

2. Move the rmtree part as part of the bsub command submitted and it works 
fine.

3. Do a touch on "DirectoryNeeded" just before reading it and everything 
works.


Almost as if the contents of the dir are still in the
rmtree DirectoryNeeded state and hasnt been updated.

Ofcourse, all of this works on Solaris.

Many thanks,

Please let me know if you would like to see more information.

Thanks,
Ram.

Re: Autofs 4.1.4_beta2 + NetApp Filer + missing dir

[mmarion]

On  1 Mar, devnull@adc.idt.com wrote:

> This is an extension to the "unacceptable bug in autofs kernel module" 
> sent by mmarion at qualcomm

I didn't send that.. someone else did, but I did comment that I thought we
were seeing the same thing.  I try to be more subtle when talking about a 
bug.. especially when I know I can't code well enough to fix it. :)

> /usr/sbin/automount -V
> Linux automount version 4.1.4_beta2
> (4.1.2 didnt work either)
> 
> We are submitting jobs to machines via LSF, the data being stored on 
> NetApp filers, running OnTap 6.5.2

Hmm.. read that you're still seeing the issue, are you using direct map or
program map to mount?  Ghosting?  Our project data is all on netapp filers 
as well, not sure on the ontap version.

I just brought up 2 hosts (1 i386 suse-9.2 and 1 amd64 sles9, both kernel 
2.6.5-7.147-smp which has the latest autofs4/ghosting patches already) and
installed 4.1.4_beta2.  I'm now doing direct mounts with ghosting.. and it
appears to be working like a charm.. interactively at least.

Next I want to push into our LSF farm and have users hit it with jobs to see
if the direct map vs program map and/or ghosting help with the jobs that say
"file not found" when it's clearly there.  

That problem almost seems like the programs expect the mount to be atomic, and
error out before automount has a time to do it's mount.. even though it's
sub-second in it's action.

-- 
Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com
"One of the best ways to get yourself a reputation as a dangerous
 citizen these days is to go about repeating the very phrases which
 our founding fathers used in the great struggle for independence."
                            -- Charles Austin Beard (1874-1948)

Re: Autofs 4.1.4_beta2 + NetApp Filer + missing dir

[ramana]

mmarion@qualcomm.com wrote:
> On  1 Mar, devnull@adc.idt.com wrote:
> 
> 
>>This is an extension to the "unacceptable bug in autofs kernel module" 
>>sent by mmarion at qualcomm
> 
> 
> I didn't send that.. someone else did, but I did comment that I thought we
> were seeing the same thing.  I try to be more subtle when talking about a 
> bug.. especially when I know I can't code well enough to fix it. :)
> 
> 
I try to be this way.


> That problem almost seems like the programs expect the mount to be atomic, and
> error out before automount has a time to do it's mount.. even though it's
> sub-second in it's action.
> 

No.



Regards
ramana

-- 
http://www.intraperson.com