litespeed stuck

[jackc]

i'm a new litespeed user, just got litespeed installed on some servers replacing apache.
tonight i notice litespeed on one of the servers got stuck somehow, unable to access litespeed admin panel, unable to open up websites, while server load is very low and not much network activity. i checked everything and can't find out why it's stuck, finally i had to kill the litespeed process and restart litespeed, then everything works fine again.
is there a way to prevent this problem? except use a http service monitor...

 

[jackc]

it seems litespeed got stuck when directadmin doing daily log rotation and backup, not sure why just this server has this problem.

 

[mistwang]

If I/O wait is very high, it may affect the lshttpd, please try "strace -tt -T -p <pid_lshttpd>" see what is going on.

AIO in 4.1 will solve this issue.

 

[jackc]

this problem start happening on another server, with very low i/o wait.
strace result is too long.

 

[NiteWave]

it seems litespeed got stuck when directadmin doing daily log rotation and backup, not sure why just this server has this problem.

it may relate to the log rotation.

for example:

/etc/logrotate.d>cat syslog 
/var/log/kernel /var/log/messages {
    sharedscripts
    postrotate
        /bin/kill -HUP `cat /var/run/syslogd.pid 2> /dev/null` 2> /dev/null || true
    endscript
}

for lsws log, it's not necessary to execute the postrotate command:

        /bin/kill -HUP `cat /var/run/syslogd.pid 2> /dev/null` 2> /dev/null || true

can you find out the log rotate script of DirectAdmin, and try to change it.

update: following example is better than syslogd:

/etc/logrotate.d>cat httpd
/var/log/httpd/*log {
    missingok
    notifempty
    sharedscripts
    postrotate
        /sbin/service httpd reload > /dev/null 2>/dev/null || true
    endscript
}

 

[jackc]

it's possible caused by high i/o when the system was cleaning /tmp
really hope the next release fix this problem because it's annoying.

 

[NiteWave]

it's possible caused by high i/o when the system was cleaning /tmp
really hope the next release fix this problem because it's annoying.

/tmp/lshttpd is used by lshttpd, can't be deleted when lsws is running.

 

[jackc]

no, it was cleaning php opcode caches.

 

[NiteWave]

no, it was cleaning php opcode caches.

I used to use APC, and never experienced "lsws stuck" when cleaning APC cache. Maybe you can try the built-in clean-up function in the opcode cache you're using if it exists, instead of shell command rm.

 

[mistwang]

If there are enough memory on your server, you can move php opcode cache to under /dev/shm.

 

[jackc]

i know. but there could be other things causing temporary high i/o, too much randomness, so i just want to see lshttpd don't get stuck if that happens.

 

[mistwang]

4.1 release with AIO will do much better in this regard.