Michael Kublin has posted comments on this change.
Change subject: core: ReconstructMasterDomain - refresh storage pool before
connect(#841525)
......................................................................
Patch Set 3: Do not submit
(2 inline comments)
First of all the patch is bad, it introduces a errors and also make code more
unreadable. Yes, if a code is bad it is no important how many comments were
added, code still will be bad. Also, please explain why these should be done :
"refresh storage pool command should be sent before connect storage pool
command to the hosts" , these is required behaviour for version 3.0 and there
are some changes in 3.1 that such change should be done? What about 3.0
Also as I know refresh should be done also after connectstoragepool or these is
no need any more?
And except all of these you are fixing a wrong bug.
1. From logs a connect usually successes if for that host was send disconnect,
as far I remember the connect should usually success and there was request not
to send a disconnect to host side.
2. The host which actually successes was not chosen as SPM and that is a main
cause for reconstruct loop. These what should be fixed, as SPM should be chosen
a host which successes during a whole flow. I think it obvious that there are
usually can be a case when one of the hosts will fail and it will be chosen as
SPM and these will cause again for reconstruct loop.
-2 because a solution is wrong for the bug and will introduce a new regressions
....................................................
File
backend/manager/modules/bll/src/main/java/org/ovirt/engine/core/bll/storage/ReconstructMasterDomainCommand.java
Line 219: try {
Line 220: VDSReturnValue returnValue = null;
Line 221: // refresh command shouldn't be executed for the
reconstruct initiating host
Line 222: if (!_isLastMaster && commandSucceeded &&
!getVds().equals(vds)) {
Line 223: returnValue =
executeRefreshStoragePoolCommand(vds);
These condition will never will be true.
You know that if refresh storage pool will failed an exception will be thrown.
In order to understand why u can look on code or at least during verifying your
code.
Line 224: if (!returnValue.getSucceeded()) {
Line 225: log.errorFormat("Post reconstruct actions
(refreshPool) did not complete on host {0} in the pool. error {1}",
Line 226: vds.getId(),
Line 227:
returnValue.getVdsError().getMessage());
Line 228: }
Line 229: }
Line 230: // the host that executed the reconstruct command
doesn't perform refresh
Line 231: // so even if refresh command wasn't executed in
the loop (returnValue == null), we need to
Line 232: // connect to the storage pool.
Please explain to me when you will get returnValue == null, if you will get it
is a bug and should be fixed, and not by adding some condition
Line 233: if (returnValue == null ||
returnValue.getSucceeded()) {
Line 234: returnValue =
executeConnectStoragePoolCommand(vds);
Line 235: if (!returnValue.getSucceeded()) {
Line 236: log.errorFormat("Post reconstruct actions
(connectPool) did not complete on host {0} in the pool. error {1}",
--
To view, visit http://gerrit.ovirt.org/7137
To unsubscribe, visit http://gerrit.ovirt.org/settings
Gerrit-MessageType: comment
Gerrit-Change-Id: I9b1b6c32ecb1c0d3c0a9ef14beef333e442c6ccf
Gerrit-PatchSet: 3
Gerrit-Project: ovirt-engine
Gerrit-Branch: master
Gerrit-Owner: Liron Aravot <[email protected]>
Gerrit-Reviewer: Allon Mureinik <[email protected]>
Gerrit-Reviewer: Ayal Baron <[email protected]>
Gerrit-Reviewer: Daniel Erez <[email protected]>
Gerrit-Reviewer: Liron Aravot <[email protected]>
Gerrit-Reviewer: Maor Lipchuk <[email protected]>
Gerrit-Reviewer: Michael Kublin <[email protected]>
Gerrit-Reviewer: Sharad Mishra <[email protected]>
Gerrit-Reviewer: Tal Nisan <[email protected]>
_______________________________________________
Engine-patches mailing list
[email protected]
http://lists.ovirt.org/mailman/listinfo/engine-patches
