vguilleaume
commented
2 years ago I continue to work on this to understand what's happened, i went through the question of the table size of the scatter-gather. I verify the value of the table size of our Qlogic adaptor and i can confirm that the value is 1024 , higher than 256.
root@nodeum:~# cat /sys/class/scsi_host/host2/model_name
QLE2672
root@nodeum:~# cat /sys/class/scsi_host/host2/sg_tablesize
1024
root@nodeum:~# cat /sys/class/scsi_host/host3/model_name
QLE2672
root@nodeum:~# cat /sys/class/scsi_host/host3/sg_tablesize
1024Even if allow_dio is set at 0, potentially the problem we have can come from the Emulex adaptors ; we will test the default size of the adaptor and keep you informed.
piste-jp
Bug Description We would like to report an LTFS problem we are facing with one specific case which 'break' LTFS mount point with unexpected sg / drive / .... errors.
We are familiar in using LTFS solution ; and we have multiple different system using the same solution which included OpenLTFS software ; and we are facing strange experience with one specific LTFS implementation.
We execute a mount LTFS of a tape 'exemple 000043L8' with success , we start to copy files into the LTFS mount point of the tape 000043L8. This works fine , and copy can be executed during at least one hour. Very stable.
In // , we start to execute others LTFS operations such as others LTFS mount or LTFS format operations ; these operations work fine during couple of minutes. 10 / 15 Minutes.
Suddenly, something happened that 'break' all LTFS mount operations. The first LTFS mount is not accessible anymore ; we can't copy files anymore. Then we do an LTFS check - try to umount - remount and after couple of try, we can have it remounted. And after couple of minutes, we have again the same issue.
We have experiences with such kind of operations (but with others hardware components : tape library - drive - hba - .... ) ; and we have this situation in only 1 specific case.
Hardware context
To Reproduce Execute multiple LTFS operation in // into multiple drives with multiple sg devices.
Expected behavior Get back the LTFS mounted operations stable.
Trace - Log We have different logs of the LTFS operation and we have detected some strange issues. We had a look into others opened issues and more specially the one related to the HBA Emulex, ...
In the
var/log/messages, we detect LTFS issues such asFailed to execute SG_IO ioctl, opcode = 0a (12).but also message in from the kernel such askernel: st 18:0:0:0: Mode parameters changedandst 18:0:0:0: [sg6] tag#0 Add. Sense: Rounded parameterThe
allow_diois at 0 ; and we don't have used this option--enable-buggy-ifsExtract of the logs , i tried to clean the log in keeping the 'most interesting' information (i have others similar logs i can shared) :
I have no idea about what's happened, thanks for your further support,
Software Version