On 06/01/2011 09:33 PM, Sharad Mishra wrote:
# HG changeset patch
# User Sharad Mishra<snmishra(a)us.ibm.com>
# Date 1306954834 25200
# Node ID 7ecdbe3a4e204ff52fb0c336fdd6cd9f7c80b14a
# Parent 8e7c9a8ad1f9e5bf6765c89a36ccca43f0564934
Migration was failing due to lost libvirt connection.
One of the changes made to fix libvirt connection leaks, broke VM migration. The change
has been undone. This change was not causing the leaks.
Signed-off-by: Sharad Mishra<snmishra(a)us.ibm.com>
diff -r 8e7c9a8ad1f9 -r 7ecdbe3a4e20 src/Virt_VSMigrationService.c
--- a/src/Virt_VSMigrationService.c Tue May 24 15:29:32 2011 -0300
+++ b/src/Virt_VSMigrationService.c Wed Jun 01 12:00:34 2011 -0700
@@ -1294,6 +1294,7 @@
"Completed");
raise_deleted_ind(job);
+ virConnectClose(job->conn);
free(job->domain);
free(job->ref_cn);
free(job->ref_ns);
@@ -1510,7 +1511,6 @@
out:
CMReturnData(results, (CMPIValue *)&retcode, CMPI_uint32);
- virConnectClose(job->conn);
return s;
}
+1.
On a side note, I think we should employ one rule that is already used
on libvirt development process. If a patch is trivial enough or if it
fixes a crash or a broken build, it can be pushed without a prior
acknowledge.
Cheers!
--
Eduardo de Barros Lima
Software Engineer, Open Virtualization
Linux Technology Center - IBM/Brazil
eblima(a)br.ibm.com