oversized [ports] section && 'stopped reporting'
list Jerry Yu
one of our servers have up to 6000 TIME_WAIT tcp connections from time to time. This got truncated, of course. I want to see them all. any quick way to change the upper limit w/o recompilation ? Also, for the same server, from time to time, certain checks 'stopped reporting' [ files & msgs, in particular ] for some extended period, sometimes over 35m, while other checks kept going. I wonder if the message truncation triggered by the over-sized [ports] section caused this loss. Or should I look elsewhere ? This 'reporting-stopped' symptoms occur with or without an actual PORT check configured for the clients under questions.
list Henrik Størner
▸
On Fri, Aug 04, 2006 at 09:54:04AM -0400, Jerry Yu wrote:
one of our servers have up to 6000 TIME_WAIT tcp connections from time to time. This got truncated, of course. I want to see them all. any quick way to change the upper limit w/o recompilation ?
Set the MAXMSG_CLIENT setting in hobbitserver.cfg and restart Hobbit. The default is 512 (KB) for the maximum size of a client message. If you want them all to show up on the status display, you probably also have to increase the MAXMSG_STATUS setting. The hobbitserver.cfg man-page describes these.
▸
Also, for the same server, from time to time, certain checks 'stopped reporting' [ files & msgs, in particular ] for some extended period, sometimes over 35m, while other checks kept going. I wonder if the message truncation triggered by the over-sized [ports] section caused this loss.
It could very well be due to this. The [files] and [msgs] sections of the client message appear after the network ports listing, so if the message was truncated these could be lost. Regards, Henrik
list Jerry Yu
thanks, I doubled MAXMSG_CLIENT MAXMSG_STATUS as well as MAXMSG_DATA from their default values. That seem to have eliminated the truncation line for PORTS data and oversize status/stachg messages for Hobbitd. after "/etc/init.d/hobbit stop" and 'kill -TERM' the remaining vmstat processes, I had to 'ipcrm' one last shm segment.
▸
On 8/6/06, Henrik Stoerner <user-ce4a2c883f75@xymon.invalid> wrote:On Fri, Aug 04, 2006 at 09:54:04AM -0400, Jerry Yu wrote:one of our servers have up to 6000 TIME_WAIT tcp connections from time to time. This got truncated, of course. I want to see them all. any quick way to change the upper limit w/o recompilation ?Set the MAXMSG_CLIENT setting in hobbitserver.cfg and restart Hobbit. The default is 512 (KB) for the maximum size of a client message. If you want them all to show up on the status display, you probably also have to increase the MAXMSG_STATUS setting. The hobbitserver.cfg man-page describes these.Also, for the same server, from time to time, certain checks 'stopped reporting' [ files & msgs, in particular ] for some extended period, sometimes over 35m, while other checks kept going. I wonder if the message truncation triggered by the over-sized [ports] section caused this loss.It could very well be due to this. The [files] and [msgs] sections of the client message appear after the network ports listing, so if the message was truncated these could be lost. Regards, Henrik