On 03/05/2014 05:05 AM, Bill MacAllister wrote: > >> Hmmm!!! If I make the count, from these logs, it shows that all your >> paths went faulty. >> Either, you had permanent faulty paths, or else the status of the paths >> wasn't updated because of a bug. > > That the paths are fine when using the 2.6 kernel and fail with the 3.2 > kernel argues against it being a faulty path.
For the running kernel, where you encountered this problem, that is the state of the paths. A kernel regression (SCSI or Transport) is a possible cause too. > >> And as expected, now everything is perfect. >> >> From what I look into your logs, it is hard to conclude the cause. It >> looks like a case of stale path status, which actually failed, when you >> did a write() using the mkfs command. >> If you have the resources, you may want to try the directio path checker >> to see if it provide the real status. > > How do I do that? Look at the fields path_checker and prio. They need to be set in multipath.conf. The hwtable entry might be not current. > >> Also at the time when mkfs failed, was the SCSI device responding ? I >> mean the TUR command definitely was, but what about, say, running the >> mkfs directly on the SCSI device ?? > > Do you mean using the /dev/dm-n device instead of the mapper device? > I did try that once today, but it didn't seem to make a difference. Nope. I meant the SCSI device. -- Ritesh Raj Sarraf | http://people.debian.org/~rrs Debian - The Universal Operating System
signature.asc
Description: OpenPGP digital signature
