|
I currently have a similar setup in the works, and my solution is to use
'mon' instead of ldirectord. mon can be remotely connected to, so both
machines monitor the other's mon port. If it stops responding, I will use an
ssh command to restart it. If it doesn't respond after that, I'll break the
heartbeat.
-----Original Message-----
From: Billy Quinn [mailto:bquinn@xxxxxxxxxx]
Sent: Friday, June 16, 2000 11:54 AM
To: 'lvs-users@xxxxxxxxxxxxxxxxxxxxxx'
Subject: ldirectord dying ...
Hi all,
I'm pretty new to LVS/Ultramonkey so forgive me if this question has been
answered before ( I've checked the archives of this list and didnt see an
answer ).
I've set up an architecture somewhat similar to the highly available, load
balanced, single virtual service examples in the documentation - where we
have a number of servers residing in a private network.
I'm using heartbeat/ldirectord to monitor the servers at an application
level , and it works great . However , I'm wondering what about the
situation where ldirectord dies ( say abruptly to a kill -9 ). It goes away
quietly ( without modifying the ipvs table entries ), and heartbeat
continues to function as usual.
Question is , can heartbeat be set up to respawn a process when it dies (
ala ldirectord ) , or do I need to hand roll my own solution to that . Right
now , Im thinking of having a cron job kill -0 the ldirectord process every
once and a while and restarting it if it dies. Am I missing something here ?
I've tested it , and ldirectord can be hand started , and it picks up like
nothing every happened. Is that safe to do ?
Any thoughts/ideas on this ?
Thanks,
Billy
|
