Re: Odd Failover Predicament
"Peter J. Milanese" <[email protected]>
| Newsgroups | gmane.linux.redhat.piranha |
|---|---|
| Message-ID | <[email protected]> |
Here's more debug- This is lvs02 WHILE it is not failed over, manually running the command: /usr/sbin/nanny -c -h 192.168.1.11 -p 80 -v -s 'GET / HTTP/1.0\r\n\r\n' -x HTTP -a 15 -I /sbin/ipvsadm -t 6 -w 1 -V 149.123.1.56 -M m -U ssh --lvs Mar 20 13:40:18 lvs02 nanny[1609]: Opening TCP socket to remote service port 80... Mar 20 13:40:18 lvs02 nanny[1609]: Connecting socket to remote address... Mar 20 13:40:18 lvs02 nanny[1609]: DEBUG -- Posting CONNECT poll() Mar 20 13:40:18 lvs02 nanny[1609]: Sending len=18, text="GET / HTTP/1.0^M ^M " Mar 20 13:40:18 lvs02 nanny[1609]: DEBUG -- Posting READ poll() Mar 20 13:40:18 lvs02 nanny[1609]: DEBUG -- READ poll() completed (1,1) Mar 20 13:40:18 lvs02 nanny[1609]: Posting READ I/O; expecting 4 character(s)... Mar 20 13:40:18 lvs02 nanny[1609]: DEBUG -- READ returned 4 Mar 20 13:40:18 lvs02 nanny[1609]: READ expected len=4, text="HTTP" Mar 20 13:40:18 lvs02 nanny[1609]: READ got len=4, text=HTTP Mar 20 13:40:18 lvs02 nanny[1609]: avail: 1 active: 0: count: 2 Mar 20 13:40:18 lvs02 nanny[1609]: making 192.168.1.11:80 available This is when lvs02 takes over as lvs, using the same command Mar 20 13:47:32 lvs02 nanny[1713]: Opening TCP socket to remote service port 80... Mar 20 13:47:32 lvs02 nanny[1713]: Connecting socket to remote address... Mar 20 13:47:32 lvs02 nanny[1713]: DEBUG -- Posting CONNECT poll() Mar 20 13:47:32 lvs02 nanny[1713]: Sending len=18, text="GET / HTTP/1.0^M ^M " Mar 20 13:47:32 lvs02 nanny[1713]: DEBUG -- Posting READ poll() Mar 20 13:47:38 lvs02 nanny[1713]: READ to 192.168.1.11:80 timed out Mar 20 13:47:38 lvs02 nanny[1713]: avail: 0 active: 0: count: 2 Mar 20 13:47:38 lvs02 nanny[1713]: No service active & available... I do not know how realistic this would be, but the primary backend interfaces are what nanny uses to hit the real servers (not the faux interface, in my case 192.168.1.254). lvs01 is working like a charm, and will assume the activity if lvs02 is brought down. The problem is defnately isolated to backend access... Please advise ! Thanks Peter J. Milanese On Wed, 19 Mar 2003, Peter J. Milanese wrote: > Hello folks- > > I've gotten lvs stable with the exception of one thing... > > The Backup LVS box cannot do an HTTP get AFTER failing over. > > It can prior to the failover, which leads me to believe that > it may be an apache thing.. Running 2.0.44... FTP is not > hit with Nanny either though... > > Confirming this, coldfusion (yes.. cf) is hit fine on failover. > > In short (after failing over): > > Cold Fusion's web server works . > Apache and FTP do not. > > Trust is established between ALL servers (ssh and http). > > Real Server can hit external sites (so routing works) > > > All else seems to failover correctly. New interfaces come up, > nanny starts, nanny gets bad results and it cannot enable the real > servers because of this. It appears NOT to be an lvs config > issue, although I'm not completely convinced since I haven't > actually found any problem. > > Anyone know where I may have mussed up? > > Thanks- > > Peter >