AW: [OpenAFS] OpenAFS crashes when using it on overlayfs

2016-10-06 Thread Richter, Michael
That was a good point. I worked for weeks to get this working and now it's that 
simple. I started with AuFS, where OpenAFS client didn't crash on boot but also 
didn't work on login.

OK, now it's working with memcache. That's good for me as I don't want anything 
to be written onto harddisc. I'll test it some more but it's looking good.

Big thanks to you two!

Mit freundlichen Grüßen
Michael Richter

-- 
Michael Richter

Technische Universität Berlin
Universitätsbibliothek
IT-Service

Fasanenstraße 88, 10623 Berlin
Telefon: +49 (0)30 314-76310
m.rich...@tu-berlin.de

www.ub.tu-berlin.de


-Ursprüngliche Nachricht-
Von: openafs-info-ad...@openafs.org [mailto:openafs-info-ad...@openafs.org] Im 
Auftrag von Karl Behler
Gesendet: Donnerstag, 6. Oktober 2016 14:46
An: openafs-info@openafs.org
Betreff: Re: [OpenAFS] OpenAFS crashes when using it on overlayfs

Hi,

I guess Volkmar is quite right suspecting AFS cache file system issues 
as the cause of your problem.

Sorry can't hellp with that, but another cache configuration might help.

We made good experiences to put the AFS cache on a tmpfs in swap which 
finally is held by the OS in memory.
This requires an entry in /etc/vfstab (on Linux it's /etc/fstab) looking 
like follows:

#device device  mount   FS  fsck
mount   mount
#to mount   to fsck point   typepass
at boot options
#
...
swap-   /var/afs-cache  tmpfs   -   
yes size=16384m

This kind of AFS configuration works like a disk cache (no need to 
modify your afsd startup parameters),
but provides speed like a memcache because tmpfs is managed in memory as 
long as some is available.
I should say our OS environment is Solaris 10 and 11. However, should 
work on Linuxes very similar.

Regards,

Karl

On 06.10.16 13:31, Volkmar Glauche wrote:
> Hi Michael,
>
> I don't know what is going on in detail here. Since the kernel log says
> something about AFS cache scan before the crash, there might be some
> issue with your cache partition. If your overlay covers the entire
> client filesystem, does that include AFS cache as well? Would it be
> possible to either mount a non-overlay partition for the cache (disk or
> ram)? Another workaround might be to use a memory cache instead of a
> disk cache.
>
> Best,
>
> Volkmar
>
> Am Donnerstag, den 06.10.2016, 09:20 + schrieb Richter, Michael:
>> Hi,
>>   
>> I’m using Ubuntu 12.04 on ~240 computers for years with this
>> configuration without problems. I’m now trying to migrate to Ubuntu
>> 16.04. Ubuntu is booted with readonly filesystem. A initramfs script
>> sets an union fs on top to bring write access. Nearly the same as on
>> every live CD.
>>   
>> On top there is the OpenAFS client. Which connects the users home
>> share on log on. But the OpenAFS client crashes on system boot.
>>   
>> If I disable OpenAFS completely and use accounts from another domain
>> without OpenAFS share, everything works. If I don’t use the overlay
>> and mount the drive read/write OpenAFS is working too. Only the
>> combination of OpenAFS and OverlayFS (or AuFS) is not working.
>>   
>> I also tried Ubuntu‘s Upstream Kernel and the OpenAFS client from
>> OpenAFS website. Nothing helped.
>>   
>> This is what I do in initramfs‘ init-bottom script:
>>   
>> ~~~
>> ro_mount_point="${rootmnt%/}.ro"
>> rw_mount_point="${rootmnt%/}.rw"
>>   
>> echo  Create mount points for the read-only and read/write layers:
>> mkdir "${ro_mount_point}" "${rw_mount_point}"
>>   
>> echo  Move the already-mounted root filesystem to the ro mount point:
>> mount --move "${rootmnt}" "${ro_mount_point}"
>>   
>> echo  Mount the read/write filesystem:
>> modprobe -r zram
>> modprobe zram num_devices=4
>> echo 1 > /sys/block/zram0/reset
>> echo lz4 > /sys/block/zram0/comp_algorithm
>> echo 2 > /sys/block/zram0/max_comp_streams
>> echo 3G > /sys/block/zram0/disksize
>> mke2fs /dev/zram0
>> mount -t ext2 /dev/zram0 "${rw_mount_point}"
>> # also tried tmpfs here
>> #mount -t tmpfs -o rw,noatime,mode=755 root.rw "${rw_mount_point}"
>> mkdir "${rw_mount_point}/upper"
>> mkdir "${rw_mount_point}/work"
>>   
>> echo  Mount the union:
>> mount -t overlay -o
>> "lowerdir=${ro_mount_point},upperdir=${rw_mount_point}/upper,workdir=
>> ${rw_mount_point}/work" overlay "${rootmnt}"
>>   
>> mkdir "${rootmnt

Re: [OpenAFS] OpenAFS crashes when using it on overlayfs

2016-10-06 Thread Karl Behler

Hi,

I guess Volkmar is quite right suspecting AFS cache file system issues 
as the cause of your problem.


Sorry can't hellp with that, but another cache configuration might help.

We made good experiences to put the AFS cache on a tmpfs in swap which 
finally is held by the OS in memory.
This requires an entry in /etc/vfstab (on Linux it's /etc/fstab) looking 
like follows:


#device device  mount   FS  fsck
mount   mount
#to mount   to fsck point   typepass
at boot options
#
...
swap-   /var/afs-cache  tmpfs   -   
yes size=16384m

This kind of AFS configuration works like a disk cache (no need to 
modify your afsd startup parameters),
but provides speed like a memcache because tmpfs is managed in memory as 
long as some is available.
I should say our OS environment is Solaris 10 and 11. However, should 
work on Linuxes very similar.


Regards,

Karl

On 06.10.16 13:31, Volkmar Glauche wrote:

Hi Michael,

I don't know what is going on in detail here. Since the kernel log says
something about AFS cache scan before the crash, there might be some
issue with your cache partition. If your overlay covers the entire
client filesystem, does that include AFS cache as well? Would it be
possible to either mount a non-overlay partition for the cache (disk or
ram)? Another workaround might be to use a memory cache instead of a
disk cache.

Best,

Volkmar

Am Donnerstag, den 06.10.2016, 09:20 + schrieb Richter, Michael:

Hi,
  
I’m using Ubuntu 12.04 on ~240 computers for years with this

configuration without problems. I’m now trying to migrate to Ubuntu
16.04. Ubuntu is booted with readonly filesystem. A initramfs script
sets an union fs on top to bring write access. Nearly the same as on
every live CD.
  
On top there is the OpenAFS client. Which connects the users home

share on log on. But the OpenAFS client crashes on system boot.
  
If I disable OpenAFS completely and use accounts from another domain

without OpenAFS share, everything works. If I don’t use the overlay
and mount the drive read/write OpenAFS is working too. Only the
combination of OpenAFS and OverlayFS (or AuFS) is not working.
  
I also tried Ubuntu‘s Upstream Kernel and the OpenAFS client from

OpenAFS website. Nothing helped.
  
This is what I do in initramfs‘ init-bottom script:
  
~~~

ro_mount_point="${rootmnt%/}.ro"
rw_mount_point="${rootmnt%/}.rw"
  
echo  Create mount points for the read-only and read/write layers:

mkdir "${ro_mount_point}" "${rw_mount_point}"
  
echo  Move the already-mounted root filesystem to the ro mount point:

mount --move "${rootmnt}" "${ro_mount_point}"
  
echo  Mount the read/write filesystem:

modprobe -r zram
modprobe zram num_devices=4
echo 1 > /sys/block/zram0/reset
echo lz4 > /sys/block/zram0/comp_algorithm
echo 2 > /sys/block/zram0/max_comp_streams
echo 3G > /sys/block/zram0/disksize
mke2fs /dev/zram0
mount -t ext2 /dev/zram0 "${rw_mount_point}"
# also tried tmpfs here
#mount -t tmpfs -o rw,noatime,mode=755 root.rw "${rw_mount_point}"
mkdir "${rw_mount_point}/upper"
mkdir "${rw_mount_point}/work"
  
echo  Mount the union:

mount -t overlay -o
"lowerdir=${ro_mount_point},upperdir=${rw_mount_point}/upper,workdir=
${rw_mount_point}/work" overlay "${rootmnt}"
  
mkdir "${rootmnt}/ro" "${rootmnt}/rw"

mount --bind "${ro_mount_point}" "${rootmnt}/ro"
mount --bind "${rw_mount_point}" "${rootmnt}/rw"
~~~
  
I also attached the crash part from kern.log.
  
Do someone know how to get this working? Is it a bug? Am I doing

something wrong?
  
Michael Richter
  
--

Michael Richter
  
Technische Universität Berlin

Universitätsbibliothek
IT-Service
  
Fasanenstraße 88, 10623 Berlin

Telefon: +49 (0)30 314-76310
m.rich...@tu-berlin.de
  
www.ub.tu-berlin.de

  �����/��&j)b�   
b�өzpK��~~��m�b��i?���+-�w���^)�



--
Dr. Karl Behler 
CODAC & IT services ASDEX Upgrade
phon +49 89 3299-1351 fax 3299-961351

___
OpenAFS-info mailing list
OpenAFS-info@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-info


Re: [OpenAFS] OpenAFS crashes when using it on overlayfs

2016-10-06 Thread Volkmar Glauche
Hi Michael,

I don't know what is going on in detail here. Since the kernel log says
something about AFS cache scan before the crash, there might be some
issue with your cache partition. If your overlay covers the entire
client filesystem, does that include AFS cache as well? Would it be
possible to either mount a non-overlay partition for the cache (disk or
ram)? Another workaround might be to use a memory cache instead of a
disk cache.

Best,

Volkmar

Am Donnerstag, den 06.10.2016, 09:20 + schrieb Richter, Michael:
> Hi,
>  
> I’m using Ubuntu 12.04 on ~240 computers for years with this
> configuration without problems. I’m now trying to migrate to Ubuntu
> 16.04. Ubuntu is booted with readonly filesystem. A initramfs script
> sets an union fs on top to bring write access. Nearly the same as on
> every live CD.
>  
> On top there is the OpenAFS client. Which connects the users home
> share on log on. But the OpenAFS client crashes on system boot.
>  
> If I disable OpenAFS completely and use accounts from another domain
> without OpenAFS share, everything works. If I don’t use the overlay
> and mount the drive read/write OpenAFS is working too. Only the
> combination of OpenAFS and OverlayFS (or AuFS) is not working.
>  
> I also tried Ubuntu‘s Upstream Kernel and the OpenAFS client from
> OpenAFS website. Nothing helped.
>  
> This is what I do in initramfs‘ init-bottom script:
>  
> ~~~
> ro_mount_point="${rootmnt%/}.ro"
> rw_mount_point="${rootmnt%/}.rw"
>  
> echo  Create mount points for the read-only and read/write layers:
> mkdir "${ro_mount_point}" "${rw_mount_point}"
>  
> echo  Move the already-mounted root filesystem to the ro mount point:
> mount --move "${rootmnt}" "${ro_mount_point}"
>  
> echo  Mount the read/write filesystem:
> modprobe -r zram
> modprobe zram num_devices=4
> echo 1 > /sys/block/zram0/reset
> echo lz4 > /sys/block/zram0/comp_algorithm
> echo 2 > /sys/block/zram0/max_comp_streams
> echo 3G > /sys/block/zram0/disksize
> mke2fs /dev/zram0
> mount -t ext2 /dev/zram0 "${rw_mount_point}"
> # also tried tmpfs here
> #mount -t tmpfs -o rw,noatime,mode=755 root.rw "${rw_mount_point}"
> mkdir "${rw_mount_point}/upper"
> mkdir "${rw_mount_point}/work"
>  
> echo  Mount the union:
> mount -t overlay -o
> "lowerdir=${ro_mount_point},upperdir=${rw_mount_point}/upper,workdir=
> ${rw_mount_point}/work" overlay "${rootmnt}"
>  
> mkdir "${rootmnt}/ro" "${rootmnt}/rw"
> mount --bind "${ro_mount_point}" "${rootmnt}/ro"
> mount --bind "${rw_mount_point}" "${rootmnt}/rw"
> ~~~
>  
> I also attached the crash part from kern.log.
>  
> Do someone know how to get this working? Is it a bug? Am I doing
> something wrong?
>  
> Michael Richter
>  
> --
> Michael Richter
>  
> Technische Universität Berlin
> Universitätsbibliothek
> IT-Service
>  
> Fasanenstraße 88, 10623 Berlin
> Telefon: +49 (0)30 314-76310
> m.rich...@tu-berlin.de
>  
> www.ub.tu-berlin.de
>  

[OpenAFS] OpenAFS crashes when using it on overlayfs

2016-10-06 Thread Richter, Michael
Hi,

I'm using Ubuntu 12.04 on ~240 computers for years with this configuration 
without problems. I'm now trying to migrate to Ubuntu 16.04. Ubuntu is booted 
with readonly filesystem. A initramfs script sets an union fs on top to bring 
write access. Nearly the same as on every live CD.

On top there is the OpenAFS client. Which connects the users home share on log 
on. But the OpenAFS client crashes on system boot.

If I disable OpenAFS completely and use accounts from another domain without 
OpenAFS share, everything works. If I don't use the overlay and mount the drive 
read/write OpenAFS is working too. Only the combination of OpenAFS and 
OverlayFS (or AuFS) is not working.

I also tried Ubuntu's Upstream Kernel and the OpenAFS client from OpenAFS 
website. Nothing helped.

This is what I do in initramfs' init-bottom script:

~~~
ro_mount_point="${rootmnt%/}.ro"
rw_mount_point="${rootmnt%/}.rw"

echo  Create mount points for the read-only and read/write layers:
mkdir "${ro_mount_point}" "${rw_mount_point}"

echo  Move the already-mounted root filesystem to the ro mount point:
mount --move "${rootmnt}" "${ro_mount_point}"

echo  Mount the read/write filesystem:
modprobe -r zram
modprobe zram num_devices=4
echo 1 > /sys/block/zram0/reset
echo lz4 > /sys/block/zram0/comp_algorithm
echo 2 > /sys/block/zram0/max_comp_streams
echo 3G > /sys/block/zram0/disksize
mke2fs /dev/zram0
mount -t ext2 /dev/zram0 "${rw_mount_point}"
# also tried tmpfs here
#mount -t tmpfs -o rw,noatime,mode=755 root.rw "${rw_mount_point}"
mkdir "${rw_mount_point}/upper"
mkdir "${rw_mount_point}/work"

echo  Mount the union:
mount -t overlay -o 
"lowerdir=${ro_mount_point},upperdir=${rw_mount_point}/upper,workdir=${rw_mount_point}/work"
 overlay "${rootmnt}"

mkdir "${rootmnt}/ro" "${rootmnt}/rw"
mount --bind "${ro_mount_point}" "${rootmnt}/ro"
mount --bind "${rw_mount_point}" "${rootmnt}/rw"
~~~

I also attached the crash part from kern.log.

Do someone know how to get this working? Is it a bug? Am I doing something 
wrong?

Michael Richter

--
Michael Richter

Technische Universität Berlin
Universitätsbibliothek
IT-Service

Fasanenstraße 88, 10623 Berlin
Telefon: +49 (0)30 314-76310
m.rich...@tu-berlin.de

www.ub.tu-berlin.de



kern.log
Description: kern.log