|
I have used LVS on the 2.4 kernel tree for quite some time with
keepalived, the product has worked great. Thank you to the LVS team!
Glad to hear that.
I recently switched to new 64bit xeons with kernel 2.6.17.6 and I am now
having some strange issues with LVS.
Currently I am getting flooded in /var/log/messages with kernel: IPVS:
ip_vs_wrr_schedule(): no available servers.
Means: either no RS are in the service pool or all RS are quiesced.
However, running ipvsadm reports I have all of my real servers loaded
into IPVS.
I have 48 entries in the ipvsadm table, I am pasting one of them here:
TCP imap.mydomain.net:imap wrr persistent 300
-> localhost.localdomain:imap Local 0 0 0
-> 10.0.5.56:imap Route 0 0 0
Yes, but they all have weight zero (quiesced) and are thus no being
elected for service forwarding by the director. Something (health
check?) must have zeroed them.
IPVS is also not functioning as you can imagine since it seems to think
it has no real servers.
I have seem some minor changes between 2.6.17.6 and 2.6.19.2 but nothing
I can see that relates to any bugfixes, so by upgrading I do not think
it would solve this problem. Have I overlooked something?
Maybe something you log files?
I have tried shutting off the real servers and let keepalived fail out
of every service which results in the real servers being removed from
ipvsadm. ...I have then re-activated the real servers which adds the
real servers back in correctly to ipvsadm. This did not fix the problem
but keepalived appeared to be operating correctly. I have fixed the
problem by killing keepalived and starting it back up. Perhaps this
unloads and loads the ipvs kernel module?
Unlikely, do the keepalived log files tell you something peculiar?
I am also still working on the steps to reproducing this problem.
Has anyone seen anything like this?
Tons of times :).
Best regards,
Roberto Nibali, ratz
--
echo
'[q]sa[ln0=aln256%Pln256/snlbx]sb3135071790101768542287578439snlbxq' | dc
| 