It's okay, we all have meetings.
The "compute_readdir_cookie failed" isn't an error; it just indicates
that your FSAL (Gluster in this case) doesn't support cookie
computation. Only FSAL_RGW does that right now, to it's not an issue.
It just means that new dirents become detached, rather than being
inserted into the correct chunk.
A clean run with a clean log would be really useful, if you can get it.
Also, if you could do a packet dump on the run as well, that would also
be useful, but not strictly necessary.
Daniel
On 8/12/19 1:05 PM, Erik Jacobson wrote:
I could not find 'erikj' in my 250M log file (I could send if
if that
would help, it should compress down well).
I'm fully game to stop ganesha, clear the logs, and start ganesha for a
run that would hopefully have a smaller log file but capture everything.
I'm game to do anything that helps! I can also do a test without overlay
if that helps.
Going along your idea, I did see some snips above where the failure happened, like this:
(Sorry for being tardy; does not represent lack of interest, was caught
in meetings).
So happy for the help thnak you -
[root@leader1 ganesha]# grep home ganesha.log | grep -v oscar
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22] nfs3_lookup
:NFS3 :DEBUG :REQUEST PROCESSING: Calling nfs_Lookup handle: File Handle V3: Len=40
4300000a20c81d9f835a974ac8a16529d95910db736659947ae4a3413992521809ab9045b2000000 name:
home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22] mdc_lookup
:NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Lookup home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdc_try_get_cached :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Look in cache home, trust
content yes
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_avl_lookup :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Lookup home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_avl_lookup :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: entry not found home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdc_try_get_cached :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: mdcache_avl_lookup home
failed trust negative no
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22] mdc_lookup
:NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Try again home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdc_try_get_cached :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Look in cache home, trust
content yes
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_avl_lookup :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Lookup home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_avl_lookup :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: entry not found home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdc_try_get_cached :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: mdcache_avl_lookup home
failed trust negative no
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22] mdc_lookup
:NFS READDIR :DEBUG :NFS READDIR: DEBUG: Cache Miss detected for home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_alloc_and_check_handle :INODE :F_DBG :lookup Created entry 0x7fd220037460 FSAL
GLUSTER for home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_dirent_add :INODE :F_DBG :Add dir entry home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_avl_insert :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Insert dir entry
0x7fd220033f90 home
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
mdcache_avl_insert :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Inserted dirent home with
ckey hk=63dd25586f8d2412 fsal=0x7fd3a722f5e0
key=0xc81d9f835a974ac8a16529d95910db738e97d75725094c3db72d3497b5ac887f
12/08/2019 08:52:54 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_22]
place_new_dirent :NFS READDIR :F_DBG :NFS READDIR: FULLDEBUG: Could not add home to chunk
for directory 0x7fd2340056d0, compute_readdir_cookie failed
On Mon, Aug 12, 2019 at 11:39:28AM -0400, Daniel Gryniewicz wrote:
> On 8/12/19 10:00 AM, Erik Jacobson wrote:
>>> Okay, could you get a FULL_DEBUG log of the issue happening and post it?
>>
>> I am happy to get you snips from the various points if that's
>> interesting - like when the NFS mount happens, when the overlay happens.
>>
>> What I pasted below is the output in FULL_DBEUG for all components when
>> I issued the failing su command. First the output of the command, then
>> the log snip that came out at the time the command below failed. I ran
>> this from our "miniroot" environment, chroot'd in to the target
root.
>> That takes out the complexity of systemd and a full bootup. 'erikj' is
>> an account I created in the image so it exists in the nfs underdir.
>> Would you like output from the same command working when 'overlay' is
>> not in the picture? I'm so grateful for your help. I'll try to learn
>> about this output but it's overwhelming to me at first glance. :)
>>
>>
>> su: failsh-4.2# su - erikj
>> Last login: Mon Aug 12 08:53:07 CDT 2019
>> su: warning: cannot change directory to /home/erikj: Operation not supported
>> su: failed to execute /bin/bash: Operation not supported
>>
>
> So, everything in that log snippet succeeded, except a lookup on /root,
> which returned ENOENT. In particular, there's no attempt to access /home or
> /home/erikj in that snippet. Can you look for the first occurrence of erikj
> in the log, and send a snippet around that? You're looking for a line like
> this, but with "erikj" instead of "mail":
>
> 12/08/2019 08:54:45 : epoch 5d516e01 : leader1 : ganesha.nfsd-29407[svc_14]
> nfs3_lookup :NFS3 :DEBUG :REQUEST PROCESSING: Calling nfs_Lookup handle:
> File Handle V3: Len=40
4300000a20c81d9f835a974ac8a16529d95910db73b6306d8b535a493c89ec7727316bc227000000
> name: mail
>
> Daniel
>