Hello Fanis, My experience with EMC arrays is almost none. But still lets give it a shot. More below...
On Friday 15 November 2013 07:56 PM, Fanis wrote: > Package: multipath-tools > Version: 0.4.9+git0.4dfdaf2b-7~deb7u2 > Severity: important > > Dear Maintainer, > I am using a fairly old FC HBA (QLA2342) to connect to a SAN with > two fibers, so the server receives four (4) paths for every LUN. > The SAN storage is a EMC VNX 5300 and I offer one 1 TB LUN to the server. > The server is a clean debian wheezy install, latest updates. > The multipath-tools included with debian wheezy hangs when a path is missing. > Also the I/O is paused (probably queued), which probably means that > also multipathd is paused. The multipath queuing behavior only triggers if you are using the queue_if_no_path feature. Otherwise, after certain timeouts, multipath will propagate the SCSI errors to the layers above it. > I have tried every possible multipath configuration in the last 2 days and > the problem continued. > > The following is my multipath setup > > # multipath -ll > mpatha (3600901604410360094b55cdda04be312) dm-0 DGC ,VRAID > size=1.0T features='1 queue_if_no_path' hwhandler='1 emc' wp=rw > |-+- policy='service-time 0' prio=4 status=enabled > | |- 4:0:0:0 sdd 8:48 active ready running > | `- 3:0:0:0 sdb 8:16 active ready running > `-+- policy='service-time 0' prio=1 status=enabled > |- 4:0:1:0 sde 8:64 active ready running > `- 3:0:1:0 sdc 8:32 active ready running > > If I disconnect a cable, thus failing 2 paths, then the multipath command > hangs after discovering the device state > > # multipath -ll -v 10 > ... > Nov 15 18:27:52 | sdd: path state = blocked > > All I/O is paused then. And the state of the devices is the following. But, at this stage, what is the output of just `multipath -ll`. -v3 and above is as good as reconfigure. If you have 2 secondary paths available, IO should ideally switch to them. > > # cat /sys/block/sd*/device/state > running > running > blocked > blocked Hmmm!!! Does multipathd sense this state of the SCSI devices, in its table? In the `multipath -ll` output? > When I reconnect the cable then multipath unhangs and I/O continues. > > I have tried every possible configuration compatible with my storage and > the problem persists. > > The problem is fixed on the multipath-tools shipping with jessie and a simple > upgrade of the multipath-tools and it's deps resolves everything. Good. I recently pushed that one, checking out the latest from upstream. > After the upgrade no device state is blocked when a path is missing. > The multipath.conf below is an example since multipathing works > without even a configuration. The package version below are after the patial > upgrade. > > Thank you Give me the multipath -v3 output? And confirm what the defaults is being used, since in your case, you mention not using a .conf file. -- Ritesh Raj Sarraf | http://people.debian.org/~rrs Debian - The Universal Operating System
signature.asc
Description: OpenPGP digital signature
