11:49 p.m.
Hi.
I wanted to check on the status of ovirt 3.5.1 with CentOS 7.1. I'm
pretty sure the current 3.5.1 engine has problems with CentOS 7.1 (?),
but not sure about vdsm? I know that 3.5.2 will resolve issues with
engine...
(I'm asking because I've kickstarted CentOS 7.1 as a host, and am having
a few problems (eg. unable to talk to power management) and want to
avoid debugging if it's already known to be broken... :)
Thanks!
Jason.
12:18 a.m.
Once upon a time, Jason Keltz <jas(a)cse.yorku.ca> said:
(I'm asking because I've kickstarted CentOS 7.1 as a host,
and am
having a few problems (eg. unable to talk to power management) and
want to avoid debugging if it's already known to be broken... :)
The power management appears to be a bug between vdsm and the fence
agents. Are you using fence_ipmilan? It seems to not be seeing
options.
Otherwise, I'm running CentOS 7.1 + ovirt 3.5.1 on hosts okay.
My work-around (quick-n-dirty hack until I have more time to look and
file an appropriate bug) for fence_ipmilan is the following script in
/usr/local/sbin (I just need the "lanplus" option, adjust as needed):
########################################################################
#!/bin/sh
cat - > /tmp/ipmi.$$
echo -e '\nlanplus=1' >> /tmp/ipmi.$$
cat /tmp/ipmi.$$ | /usr/sbin/fence_ipmilan.real $@
rm -f /tmp/ipmi.$$
########################################################################
--
Chris Adams <cma(a)cmadams.net>
6:27 a.m.
On 23/04/2015 5:18 PM, Chris Adams wrote:
Once upon a time, Jason Keltz <jas(a)cse.yorku.ca> said:
> (I'm asking because I've kickstarted CentOS 7.1 as a host, and am
> having a few problems (eg. unable to talk to power management) and
> want to avoid debugging if it's already known to be broken... :)
The power management appears to be a bug between vdsm and the fence
agents. Are you using fence_ipmilan? It seems to not be seeing
options.
Otherwise, I'm running CentOS 7.1 + ovirt 3.5.1 on hosts okay.
engine as well?
I thought I'd read something about that not working
here... and I think I remember reading a blurb in the release notes for
3.5.2 about fixing compatibility with CentOS 7.1.
My work-around (quick-n-dirty hack until I have more time to look
and
file an appropriate bug) for fence_ipmilan is the following script in
/usr/local/sbin (I just need the "lanplus" option, adjust as needed):
########################################################################
#!/bin/sh
cat - > /tmp/ipmi.$$
echo -e '\nlanplus=1' >> /tmp/ipmi.$$
cat /tmp/ipmi.$$ | /usr/sbin/fence_ipmilan.real $@
rm -f /tmp/ipmi.$$
########################################################################
Yep .. It's for an idrac7... I need lanplus and a few other options as
well (which are apprently integrated into 3.5.2).. It was working
perfectly with ovirt-node, but now that I've switched from node to using
my own CentOS 7.1 install, the problem was introdued. It's not clear if
the same problem would occur if I was running CentOS 7. There were a
lot of errors in the engine log. I will look into more detail tomorrow.
Thanks Chris!
Jason.
6:40 a.m.
Once upon a time, Jason Keltz <jas(a)cse.yorku.ca> said:
>Otherwise, I'm running CentOS 7.1 + ovirt 3.5.1 on hosts
okay.
engine as well? I thought I'd read something about that not working
here... and I think I remember reading a blurb in the release notes
for 3.5.2 about fixing compatibility with CentOS 7.1.
No, my engine is on CentOS 6 (I installed it with oVirt 3.5.0 when
CentOS 7 engine support wasn't ready).
Yep .. It's for an idrac7... I need lanplus and a few other
options
as well (which are apprently integrated into 3.5.2).. It was
working perfectly with ovirt-node, but now that I've switched from
node to using my own CentOS 7.1 install, the problem was introdued.
Hmm, I also have idrac7, but I only need lanplus. It worked okay until
I upgraded the hosts to CentOS 7.1 this week. I suspect something
changed in the resource agents and either vdsm was sending something
slightly wrong that no longer works, or there's a new bug in the
resource agents (since vdsm didn't change).
--
Chris Adams <cma(a)cmadams.net>
9:47 a.m.
Hi,
On 04/24/2015 05:40 AM, Chris Adams wrote:
Hmm, I also have idrac7, but I only need lanplus. It worked okay
until
I upgraded the hosts to CentOS 7.1 this week. I suspect something
changed in the resource agents and either vdsm was sending something
slightly wrong that no longer works, or there's a new bug in the
resource agents (since vdsm didn't change).
Experienced a similar problem with an upgrade from 7.0 to 7.1 on my
pacemaker cluster.
This was fixed in fence-agents 4.0.11-11.el7_1 with RHBA-2015:0801-1[1]
for me.
Maybe you need to apply the latest fence-agents update?
Regards,
Rik
[1] https://rhn.redhat.com/errata/RHBA-2015-0801.html
--
Rik Theys
System Engineer
KU Leuven - Dept. Elektrotechniek (ESAT)
Kasteelpark Arenberg 10 bus 2440 - B-3001 Leuven-Heverlee
+32(0)16/32.11.07
----------------------------------------------------------------
<<Any errors in spelling, tact or fact are transmission errors>>
11:14 a.m.
On Fri, Apr 24, 2015 at 08:47:22AM +0200, Rik Theys wrote:
Hi,
On 04/24/2015 05:40 AM, Chris Adams wrote:
>Hmm, I also have idrac7, but I only need lanplus. It worked okay until
>I upgraded the hosts to CentOS 7.1 this week. I suspect something
>changed in the resource agents and either vdsm was sending something
>slightly wrong that no longer works, or there's a new bug in the
>resource agents (since vdsm didn't change).
Experienced a similar problem with an upgrade from 7.0 to 7.1 on my
pacemaker cluster.
This was fixed in fence-agents 4.0.11-11.el7_1 with RHBA-2015:0801-1[1] for
me.
Maybe you need to apply the latest fence-agents update?
And if it does not, please share your vdsm.log's "fenceNode" command. I
may hold hints regarding the failure.
4:05 p.m.
Once upon a time, Rik Theys <Rik.Theys(a)esat.kuleuven.be> said:
On 04/24/2015 05:40 AM, Chris Adams wrote:
>Hmm, I also have idrac7, but I only need lanplus. It worked okay until
>I upgraded the hosts to CentOS 7.1 this week. I suspect something
>changed in the resource agents and either vdsm was sending something
>slightly wrong that no longer works, or there's a new bug in the
>resource agents (since vdsm didn't change).
Experienced a similar problem with an upgrade from 7.0 to 7.1 on my
pacemaker cluster.
This was fixed in fence-agents 4.0.11-11.el7_1 with
RHBA-2015:0801-1[1] for me.
Maybe you need to apply the latest fence-agents update?
That's the build I'm running (from CentOS). When I was testing this at
the initial problem, it appeared that the "lanplus" option (the only
option I'm setting) was not being handled by fence_ipmilan. I could see
it running ipmitool with "-I lan", not "-I lanplus". My posted hack
of
a script force-added "lanplus" to the input of fence_ipmilan, and then
it worked.
I checked the vdsm log, and it is passing the correct options:
Thread-2372::DEBUG::2015-04-20 17:39:08,248::API::1209::vds::(fenceNode)
fenceNode(addr=10.0.9.19,port=,agent=ipmilan,user=root,passwd=XXXX,action=status,secure=False,options=lanplus,policy=None)
I need to do some more debugging to see the source of the problem (just
haven't had time yet).
--
Chris Adams <cma(a)cmadams.net>
9:58 p.m.
Once upon a time, Chris Adams <cma(a)cmadams.net> said:
The power management appears to be a bug between vdsm and the fence
agents. Are you using fence_ipmilan? It seems to not be seeing
options.
Late follow-up, but I finally had some time to look at this. Between
RHEL 7.0 and RHEL 7.1 (fence-agents 4.0.2 and 4.0.11), fence_ipmilan was
replaced, going from a compiled C program to a python script. In the
process, boolean option handling appears to have changed when reading
options from standard input (as vdsm passes them).
In the old version of fence_ipmilan, just setting "lanplus" in the oVirt
host power management screen as an option was sufficient, but in the new
version, "lanplus=1" is required. I changed all my hosts to use
"lanplus=1" and the power management test succeeds.
I've filed a RH bug against fence-agents, as IMHO this is a regression.
https://bugzilla.redhat.com/show_bug.cgi?id=1222098
--
Chris Adams <cma(a)cmadams.net>
3478
days inactive
3500
days old
7 comments
4 participants
participants (4)
-
Chris Adams -
Dan Kenigsberg -
Jason Keltz -
Rik Theys