otherwise the long downtime while waiting for other packages
to finish unpacking and installing can lead to problems
e.g., a dist-upgrade from Jessie to Stretch might take over
5 minutes to get from running the old version's prerm (with
stop) and the new version's postinst (with start).
minor update with buildfixes for newer systems (strech, at least with
arm), config parser bugs and some qdevice fixes, while we do not have
qdevice enabled.
oneline log between v2.4.0 and v2.4.2:
96f91f2 libvotequorum: Bump version 19e48a6 [build] Fix build on RHEL7.3 latest 52e6ae5 cfg: Prevents use of uninitialized buffer b0c850f Qnetd LMS: Fix two partition use case 0da1b74 man: mention qdevice incompatibilites in votequorum.5 49a9f72 Man: Fix corosync-qdevice-net-certutil link 87c363b Spec: Qdevice require same version of corosync f837f95 Config: Flag config uidgid entries 3e995cf cts: Make it run with pacemaker-1.13+ 71c9035 Low: totemsrp: Addition of the log. a275d88 qdevice: some more small man page fixes b6c71aa qdevice: Fix 'tie_breaker' in man page
Tested for a week now in cluster setups with mixed versions (v2.4.0
and v2.4.2) and upgrade tests.
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
Add a workaround for an ipv6 multicast kernel bug where ipv6 multicast
interfaces are rendered inert when another socket gets bound after them.
This problem got fixed somewhere between 3.10 and 3.16, so this patch
can be removed when we upgrade the kernel accordingly.
Dietmar Maurer [Wed, 24 Aug 2011 11:22:31 +0000 (13:22 +0200)]
improve download target
we now use the git tag to checkout the request version. We need to
include the .git dir inside the tar archive, because that is used
by configure to detect the version.
projects / corosync-pve.git / log