Yaniv, have you seen this in vdsm22, or is it just theoretic? If this is indeed a regression, when was it introduced?

(In reply to comment #2)
> Yaniv, have you seen this in vdsm22, or is it just theoretic? If this is indeed
> a regression, when was it introduced?

yes i've seen this in vdsm22-4.5-63.25.el5_6.

this is a regression due to we had such a bug that was verified

Bug 527096 - VDSM: will not report storage (iscsi) that works in partial state => can not see all LUNs

verified in sm33 vdsm-4.5-22

https://bugzilla.redhat.com/show_bug.cgi?id=527096#c4

Please see that it takes more the 5 minutes for repostats to update (there is heavy caching on idle systems).
Also if checking partial VGs check that the 1st PV is missing, other PV detection was never supported.

Created attachment 502227 [details]
logs

reproduced on vdsm22-4.5-63.25.el5_6

setup:

1 host
1 SD extended from two different targets: 
1. master domain was located on orion
2. domain was located in qanashead

steps:

1) blocked connectivity to domain located in qanashead

Data center remained active - no error reported in rhevm and in vdsm

2) blocked connectivity to master domain located in orion

the data center became unknow within 5 minutes. 

errors were reported by vdsm:

Thread-122::ERROR::2011-06-01 14:19:28,812::dispatcher::106::irs::Traceback (most recent call last):
  File "/usr/share/vdsm/storage/dispatcher.py", line 97, in run
    result = ctask.prepare(self.func, *args, **kwargs)
  File "/usr/share/vdsm/storage/task.py", line 1287, in prepare
    raise self.error
StoragePoolMasterNotFound: Cannot find master domain: ('4255d676-f7df-47ed-bcc4-0084bbf846af',)

full logs attached