Hide Forgot
Created attachment 482085 [details] vdsm.log * Domain creation failed for "MkfsError: Cannot create filesystem on device". * Devices were left dirty, even though they are not used. * Devices should be cleaned, so they could be re-used for future domains.
Paikov, what's the difference between this bug and "682513 - VDSM - Storage: VDSM should work with new mke2fs default of discard=enabled" ???
How do I reproduce this bug?
(In reply to comment #4) > How do I reproduce this bug? It happens every time when using the storage at Westford: http://10.16.45.61
(In reply to comment #5) > (In reply to comment #4) > > How do I reproduce this bug? > > It happens every time when using the storage at Westford: http://10.16.45.61 Can you specify which actions produced this error?
While some errors are susceptible to rollback, reliably rollbacking from any error that might arise in blockSD.create is unfortunately impossible. Hence, I claim it's a futile task in the first place. There is, however, a mechanism for handling errors exactly such as this: VDSM uses the domain tags to communicate whether the domain is ready or not (RHAT_storage_domain_UNREADY). While that tag is set the VG should not be used, and should be considered as marked for deletion. This tagging mechanism seems to work reliably for failures in mkfs, therefor this doesn't appear to be a bug.
Is a normal, usable error propagated back to the user about that VG? With some identifier of it?
Yes, the function raises a VolumeGroupUninitialized exception with the name of the VG.