I'm experiencing exactly the same problem Werner Fisher explained on DRBD-User for his 0.7.x setup: the drbd startup waits forever for a connection to the other node, ignoring the wfc-timeout and degr-wfc-timeout values specified in drbd.conf. As you can see at the above URL, it was a bug and it should have been already corrected in 0.7.x. My suspect is that the same bug propagated to the 0.8.x codebase and hasn't been corrected there. I've already contacted upstream, I'll add to this bug if I get any response. If I correctly identified this bug, drbd-0.8.4 shouldn't be used since it's b0rked and will hang your cluster forever should a node reboot with the other being down/absent/hang. Has anyone else tried drbd-0.8.4? Reproducible: Always Steps to Reproduce: 1. emerge drbd-0.8.4 2. degrade cluster (turn off a node) 3. start drbd on the remaining node Actual Results: drbd startup says wfc-timeout and degr-wfc-timeout are 0 (-> infinite) and waits forever, even if the user specified different values in drbd.conf Expected Results: should read correct values from config and honor them
The bug is fixed in drbd-8.0.5. I just copied the ebuild to drbd-8.0.5.ebuild, made digest, emerged. It compiles and works ok. It honors the [degr-]wfc-timeout values both in single resources and in common section. I made some tests and everything works ok on x86_64 hardened profile, I'm going to use it for production systems starting tomorrow and suggest a version bump in portage. As far as I'm concerned, it should also be unmasked since it's stable.
8.0.5 in portage. I'm stress testing 8.0.x from the begin of June, 8.0.3 has some stability problems, 8.0.4 has some bugs in drbdmeta that affects large storages but seems to be stable ...I'll mark 8.0.x stable when all the relevant parts seems to be stable.
