One of the server that I am working with has some infinitely running PHP process. Due to incorrect way of cron setup by the development team, it has caused the process hanging and not ended properly. According to them, these processes can be killed if still hang after 12 hours.
Any process which run in server will have start time (STIME). You can check this by using ps command. In this case, following result will appear:
$ ps aux | grep php root 1399 0.0 0.0 61188 740 pts/2 S+ 10:10 0:00 grep php user1 2697 0.0 0.0 100664 8340 ? Ss Jul04 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 5551 0.0 0.4 171052 78832 ? Ss Jun25 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 9913 0.0 0.5 174636 82392 ? Ss Jun22 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 11961 0.0 0.7 223276 131060 ? Ss May25 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 16455 0.0 0.4 171564 79420 ? Ss Jun24 0:01 /usr/local/bin/php /home/user1/cron/sync2server.php user1 17474 0.0 0.5 182060 90016 ? Ss Jun18 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 20094 0.0 0.6 206636 114588 ? Ss Jun03 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 22555 0.0 0.7 213548 121476 ? Ss May30 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 24670 0.0 0.7 214572 122320 ? Ss May30 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 28200 0.0 0.7 220204 127988 ? Ss May26 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 30832 0.0 0.4 170284 78168 ? Ss Jun25 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 30837 0.0 0.4 170114 88508 ? Ss 23:20 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php user1 30848 0.0 0.4 120439 80770 ? Ss 12:20 0:00 /usr/local/bin/php /home/user1/cron/sync2server.php |
If you see the STIME value mostly has started long time ago but it is still inside the process list. To kill all the older process which more that 24 hours, I use following command:
$ kill -9 `ps aux | grep php | grep sync2server | awk '$9 !~ /[0-9]:[0-9]/' | awk '{print $2}'` |
Explanation:
Using ps command, we grep anything related to “php” and “sync2server” which is the specific process that we want to kill. The 4th argument is checking whether column no 9 (STIME) column has value which NOT in “number:number” format. Process which starts for more than 24 hours, STIME value will contains word for example “Jun23″ or “2010″. The 5th argument is actually print the value of column no 2 which is the PID to be killed.
To kill process which less than 24 hours, you can use following script:
LIMIT=43200 #limit on seconds = 12 hours PROCESS="php" #change to process u want to grep count=`ps aux | grep $PROCESS | awk {'print $9'} | wc -l` for ((i=1;i<=$count;i++)) do ptime=`ps aux | grep $PROCESS | awk {'print $9'} | head -$i | tail -1` psec=`date "+%s" -d "$ptime"` csec=`date "+%s"` exectime=$((csec-psec)) if [ $exectime -gt $LIMIT ] then pid=`ps aux | grep $PROCESS | awk {'print $2'} | head -$i | tail -1` /bin/kill -9 $pid fi done |
Above scripts will try to find any PHP process which executed more than 12 hours and will kill it one by one. You can get the script to run as cron twice per day so it will automate your administration job.
Lets share if you have better idea. Cheers!
Related Posts
2 Responses to Linux: Kill Process based on Start Time (STIME)
Leave a Reply
Subscribe via Email
Sci/Tech – Google News- Google goes to the Galapagos Islands: Backpack-mounted Street View cameras ... - Daily Mail 24 May 2013
- "This is a win for Xbox customers" says Microsoft, emerging victorious from ... - Eurogamer.net 24 May 2013
- Nissan SA recalling old-gen Micras - Independent Online 24 May 2013
- Twitter unveils more secure login option - WBRC 24 May 2013
- Samsung Galaxy S4 creates history, becomes fastest-selling Android phone - India Today 24 May 2013
it’s what I’m looking for, but I think it’s a little off, that start time is what hr:min of the day the process started, not how many hrs:mins it has been running. So your script is killing the processes that start after 10am, not ones that have been running for more than 10 hrs.
I am aware of STIME is the time when the process started. Thats why when you refer to line:11, I have exectime=$((csec-psec)) where csec is the current time in seconds and psec is the start time in seconds. It will then compare the result with $LIMIT we set.