On 06/01/15 21:38, Nicholas A. Bellinger wrote: > Hi Daniel, > > On Mon, 2014-12-29 at 16:15 +0000, Daniel Piddock wrote: >> Hi, >> >> I was hoping somebody would be able to help with an issue that has me >> stumped. Sorry if this has been asked before but I haven't found a >> similar issue. >> >> I have Corosync+Pacemaker serving up an iSCSI target, using >> ocf::heartbeat:iSCSITarget and ocf::heartbeat:iSCSILogicalUnit. The >> backstore is an iblock of an LVM LV stored in DRBD on a local drive >> partition. >> >> /dev/sda3 - DRBD physical partition >> /dev/drbd0 - LVM PV >> /dev/VG0/iscsi1 - LVM LV iblock for iSCSI >> >> Every time that the target is removed and restored the WWID as seen by >> the initiators changes (fetched with "/lib/udev/scsi_id -g -u -d >> /dev/sdc"). I have multiple network routes so this changing of WWID >> seriously upsets multipath. >> >> For example, WWID is currently 36001405caa429dc7e474943bc6026541. >> Previously it was 360014057b86b6a48432485fb4201da9d >> >> Any pointers on how to make the WWID static? >> >> Target servers and initiators are running Debian stable with backport >> kernels. >> linux-image-3.16.0-0.bpo.4-amd64 3.16.7-ckt2-1~bpo70+1 >> targetcli 2.0rc1-2 >> lio-utils 3.1+git2.fd0b34fd-2 >> python-rtslib 2.1-2 >> >> Frustratingly I have access to an almost identical setup where the WWID >> is static. The only difference I can spot is that DRBD is on a separate >> drive without a partition table. >> > So using the legacy lio-utils CLI, an existing device unit serial is set > after tcm_node --establishdev $HBA/$DEV time using: > > tcm_node --setunitserialwithmd $HBA/$DEV $UNIT_SERIAL > > It sounds like tcm_node --block is being called incorrectly each time > during fail-over by the ocf script to configure the backend device, > instead of just the first time to generate the initial $UNIT_SERIAL. > > It would be useful to verify exactly which tcm_node CLI operations are > occurring on both setups in order to diagnose why one setup is doing the > right thing to reset an existing $UNIT_SERIAL using > --setunitserialwithmd, and the other is incorrectly using --block to > generate a new $UNIT_SERIAL each time. > > --nab Hi nab, Thank you for your response. I was fiddling about on my "broken" test box and finally noticed the cause of the issue in syslog: lrmd: [2601]: info: RA output: (iscsi.lun.test:start:stderr) /usr/lib/ocf/resource.d//heartbeat/iSCSILogicalUnit: line 56: openssl: command not found I must have missed it in all the noise (corosync/pacemaker is so chatty). Never would have spotted if I'd reduced the logging level though. Installing openssl results in a predictable wwid getting generated. Cheers, Dan -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
