|
Hi,
I have simple set up of 5 real servers using a LVS front end in the
LVS-TUN configuration. I have the basic fping mon to monitor the status of
backends. Everything works fine, until there there is a fault :(. The
backends are simple HTTP servers, and each receives requests at a rate of
100 requests/sec.
Mon correctly removes an unreachable backend node (node crash). And
dutifully adds it back when the node comes back up. and ipvsadm shows all
the 5 nodes are active in the set. But at this point, the node in question
sees only about 1-5 requests/sec???? While the others are seeing almost
N/4 requests/sec instead of N/5 where N was 500 in our case. Any ideas
why this would happen? Any gotchas? I saw the same behavior with wrr
scheduling as well.
Thanks in advance...
- Kiran
|
