I have a constant mysql process using 10-20% CPU, its very strange because the process is always present, if I restart apache and mysql and I just open the phpmyadmin or any website connecting to the DB I get this error.
With strace I have seen that the process details are:
futex(0x87326c4, 0x5 /* FUTEX_??? */, 1) = 1
futex(0x8732174, FUTEX_WAKE, 1) = 1
select(15, [13 14], NULL, NULL, NULL) = 1 (in [14])
fcntl64(14, F_SETFL, O_RDWR|O_NONBLOCK) = 0
accept(14, {sa_family=AF_FILE, path="ӿ"}, [2]) = 25
fcntl64(14, F_SETFL, O_RDWR) = 0
getsockname(25, {sa_family=AF_FILE, path="/var/run/mysql"}, [30]) = 0
fcntl64(25, F_SETFL, O_RDONLY) = 0
fcntl64(25, F_GETFL) = 0x2 (flags O_RDWR)
fcntl64(25, F_SETFL, O_RDWR|O_NONBLOCK) = 0
setsockopt(25, SOL_IP, IP_TOS, [8], 4) = -1 EOPNOTSUPP (Operation not supported)
The strace show that this piece of code is being repeated everytime that any website or app is using the database.
I have seen that the process call to the path /var/run/mysql in:
getsockname(15, {sa_family=AF_FILE, path="/var/run/mysql "}, [30]) = 0
...but in my system the path doesnt exists, its /var/run/mysqld/ and I cant find a call to "/var/run/mysql" in any file of the server (i have tryed with sudo grep -nr '/var/run/mysql' /)
Server specs:
CENTOS Enterprise 4.6 i686
Linux 2.6.9-67.0.7.ELsmp
Core2Duo E6750
PHP Version 5.2.5
MySQL 5.0.45
When i login to the root for the server and go to mysql process list i see a lot of processes that are connected to photo handling and the sate is Locked with over 500 seconds of each one.
I see there is one process that the state says: Copying to tmp table
I read online that if this is taking long it might lock up other queries causing your server to perform slow and only solution is to restart mysql servers.
Right now i have over 100 queries that say Locked and one that says Copying to tmp table....each time i have to restart the mysql services but this happens every 10 minutes as i have enough traffic to cause this problem.
I have a single mysql process that never stops and running with 10-200% CPU load: URL....
I restart mysql => process coming back I restart server => process coming back I kill process => process coming back
I have run: # mysqladmin -uadmin -p`cat /etc/psa/.psa.shadow` -i 1 processlist but there was noting that runs >10min.
Code: > SHOW FULL PROCESSLIST; +------+-------+-----------+------+---------+------+-------+-----------------------+ | Id | User | Host | db | Command | Time | State | Info | +------+-------+-----------+------+---------+------+-------+-----------------------+ | 328 | admin | localhost | psa | Sleep | 56 | | NULL | | 8110 | admin | localhost | NULL | Query | 0 | init | SHOW FULL PROCESSLIST | +------+-------+-----------+------+---------+------+-------+-----------------------+ 2 rows in set (0.00 sec) I have strace the PID
Code: # timeout 1m strace -f -c -p 5873 Process 5873 attached with 30 threads - interrupt to quit Process 10499 attached (waiting for parent) Process 10499 resumed (parent 5873 ready) Process 10502 attached (waiting for parent) Process 10502 resumed (parent 5873 ready) Process 10503 attached (waiting for parent)
[code]....
I found with google a hint for high cpu URL....I have deinstalled "health monitor" module, but that was not the reason.I use plesk 12.0.18 Update #38 with CentOS 6.6 (Final).
I'd like to put up here a base question which I hope some will have the goodwill to answer even though it might touch some business secrecies.
We're a gameserver hoster since around ten years, running also vserver products since over two years now. Renting a few Racks in Europe since some time we're a bit in a question mark how rootserver companies deal with the initial hardware costs for every new customer.
Rackspace and today specially power costs are huge cash eaters here in Europe. Dedicated Rootservers are huge space & power consumers per customer ratio. The initial Hw costs for every new rootserver customer might be covered after 4-6 months (if the machine has to be bought newly), adding the bandwidth and power costs it might take up to 8-9 months until a benefit might come in.
Is this the business normality in the rootserver market (waiting 9 months for any benefit, or counting only on the benefit of the 2nd customer using the older Hw), or are the better ways to handle those "initial" costs or keep them affordably low?
In My server i Can see lotz of /usr/local/apache/bin/httpd -k start -DSSL Process also some time it lead to crash of my server,is this caused to improper httpd.con file iam have 200-250 online users here i will give my apachi cofig file
Top Process %CPU 98.9 ./dialog --clear --backtitle Zend Optimizer Installation --title Zend Optimizer 3.2.2 --msgbox ?Welcome to the Zend Optimizer 3.2.2 Installation!??For more information regarding this procedure, please see the?Zend Optimizer Installation Guide.? 0 0
Top Process %CPU 98.8 ./dialog --clear --backtitle Zend Optimizer Installation --title Zend Optimizer 3.2.2 --msgbox ?Welcome to the Zend Optimizer 3.2.2 Installation!??For more information regarding this procedure, please see the?Zend Optimizer Installation Guide.? 0 0
My hosting company is telling me that I have getting frozen processes on my VPS that is causing apache to stop responding.
My VPS is CentOS, with 512 megs of memory and my website is a Vbulletin forum. I have several add-ons installed on vbulletin, and if one of those are causing the problem, how do I figure it out?
When http stops responding, I can access the server with putty on ssh. Running the "top" command, it shows that I still have 100+ megs of memory free and the processor is barely being used.
This has happened twice in just 2 - 3 days. This morning, I asked my hosting company what can be causing this, but no answer so far.
CentOS 4.5 final Apache 2.0.59 php 4.3.11 MySQL 4.1.20
I have a user who has been importing large sql databases and this causes the cpu to run at 80% - 90% this in turn causes high loads and sometimes causing the server to crash.
Is there a way to automatically limit or control how much the process can use or stop the process to allow the server to recover.
Just trying to put some sort of safe guard in place
Code: PID USER PRI NI SIZE RSS SHARE STAT %CPU %MEM TIME CPU COMMAND 9488 nobody 25 0 1612 1280 1108 S 21.1 0.2 619:37 1 perl bodyb3 200.101.193.42 0 5000 I also ran netstat -n and found this: Code: udp 15184 0 server.host.com:55936 ns1.datacenter.net:domain ESTABLISHED
I can't remember the name of the utility that lets you watch what a process is doing. You call it on a PID and you can see all the memory allocations, file IO, library loading, etc. that the process is doing as it happens. Anyone know what I'm thinking of?
