[Linux-ha-dev] Re: Bug#425434: Heartbeat-2 taking up all resources

[Andrew Beekhof]

On 5/22/07, Simon Horman <horms at verge.net.au> wrote:
> [ Resend as the linux-ha-dev address was wrong the first time around ]
>
> On Mon, May 21, 2007 at 06:39:47PM +0200, Federico Belvisi wrote:
> > Package: heartbeat-2
> > Version: 2.0.7-2
> >
> > Architecture: arm I'm using a Linksys NSLU2 with Debian Etch stable. I
> > tried to set up a simple failover between nslu2 and a Gentoo server
> > (i386). The only thing i want to failover is dhcp3-server (no ip
> > failover, only dhcp3-server). I installed the stable heartbeat on
> > gentoo and the stable branch heartbeat on Debian. Gentoo is primary,
> > Debian/NSLU2 is backup.  I configured all the things up but soon i
> > noticed that there were some little problems: gentoo complained that
> > the heartbeat versions were different, although some failover was
> > working. The problem was with auto_failback not working. I soon
> > realized that in a newbie fashion i installed heartbeat instead of
> > heartbeat-2 on Debian. I removed the old heartbeat, installed
> > heartbeat-2 and as soon as installation was at the end, my little
> > nslu2 was slowing down. After 1 minute i realized that the only thing
> > was to shutdown the nslu2. I mounted the nslu2 hard disk on pc and
> > performed regular maintenance and deleted symlinks in order not to let
> > heartbeat-2 start at boot. I double checked configuration files and
> > they were the same as they were when "the old" heartbeat (branch 1.)
> > was installed, and that they are the same as on Gentoo server. I
> > booted the nslu2, and all was fine. Load averages were between 0.50
> > and 1. I then decided to start heartbeat-2 so i typed
> > "/etc/init.d/heartbeat start". System started to slow down, being less
> > responsive. I typed "top" just to see what was process activity.
> > Heartbeat was using many (maybe 43%) of memory, and load averages were
> > steeply raising: in about 2 minutes load average was some like 15. I
> > waited some minutes more but the shell  wasn't updating (even if ssh
> > session was still up), so i decided to shutdown again the nslu2.
> >
> > I retried upgrading heartbeat-2 from testing repository (version
> > 2.0.8-1), but things were the same as above.
> >
> > I think heartbeat-2 on arm is doing something bad, by taking up all
> > resources (cpu, memory). This don't happen with package heartbeat.
>
> Hi Federico,
>
> I'm not sure what the minimal cpu and memory requirements for
> heartbeat-2 are, but perhaps someone the linux-ha-dev list has

2.0.8 had a bunch of memory leaks (and performance problems IIRC)
which wouldn't be helping.

We dont have an updated official release, but you could try the packages at
   http://software.opensuse.org/download/server:/ha-clustering/
(which are what paying SUSE customers will be using shortly) to see if
that makes things any better.

> some idea, I have Cced them. In the mean time, could you
> supply some information on ammount of CPU and memory your
> machine has.