03-06-2007 03:12 PM - edited 03-03-2019 04:03 PM
Hi
Is anyone aware of any isues using a 6509 switch with server running lighttpd
I am not getting anywhere with my NOC and am looking for answers independantly from them
we have 8 high load servers behind a 6509 switch connected to cogent
each server goes down every 12 or so hours. no apparent reason it just fails software on the server runs fine. including lighttpd
only a reboot seems to correct the issue, or patience it comes up again after about an hours "rest" (no scripts are rebooting the server software)
hard to quantify but I have also noted if one servers bandwidth usage increases another drops or maybe several?
I believe the router has overall bandwidth capping enabled to a limit of 4gb/s until recently we were running at that level 24/7 with all the problems occuring
I am surmising that perhaps something to do with lighttpd is causing the router to 'mismanage" the capping reducing the bwidth allowance on some ports in favour to other ports? it seems if this is the case the port with the most b/w is selected at random.
when running apache everything seems fine except apache cannot handle the number of concurrent connections and server load skyrockets to over 100 on average. with lighttpd it drops to 1-2 on average but these other issues surface.
Each server usually has 2000-4000 concurrent connections. downloading files in the range of 2-25Mbytes
I cannot give any information about the configuration of the 6509 as I am not privy to it.
any clues as to what may be happening could help me present an intelligent argument to the techs at the NOC giving them something to look for
Thanks in advance
03-12-2007 12:36 PM
The Catalyst 6500 series Content Switching Module (CSM) provides high-performance server load balancing (SLB) among groups of servers, server farms, firewalls, caches, VPN termination devices, and other network devices, based on Layer 3 as well as Layer 4 through Layer 7 packet information.
03-12-2007 04:28 PM
Thank you
The link you provided appears to be invalid
we have at least partially solved the issue which was due to lighttpd not being threaded, in essence it was i/o bound, so unrelated to the 6509
Unfortunately the 12 hour cycle of downtime has not been resolved. again as the 6509 is one of the very few common points we are still looking at it as at least part of the problem.
03-12-2007 10:28 PM
is it possible for you to post the config>?
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide