Need Suggestions to improve performance of the linux server
 

Hi All,

We are using centos in our machines but two of the linux boxes are performing very slow, while restoring the mysql database it tooks huge time to complete.

Can someone pls advice what we can do to find out the problem with the slow performance?

Hi,

Try to check if there are any errors in the mysql log file, for standard mysql-server centos package log file can be found in : /var/log/mysqld.log . Check /etc/my.cnf for more information.

Do you have performance issues only when a database restoration is performed?

checking for mysql-specific errors makes sense, but it seems like you'd notice if there were real problems. my first step is to run "vmstat 1" during the slow time. it lets you see, for instance, whether the problem is CPU (100% busy, no idle time), or memory (swapins means thrashing - swapouts are not inherently bad.) it also gives some idea of whether disk(s) are not keeping up ("wait" - processes that are blocked on IO.)

it would also be useful to know what you mean by "performing very slow" - do you mean interactive feel? or do you have some application which reports response times (such as a webserver)? it could also be that your configuration is simply inappropriate - any real analysis would need to know what you're running, and running on...

Hi

The server itself is performing slow (While doing file compression, taking database dump) and during mysql database restoration it takes more time to complete than other systems. We use the server mainly as database server.

still to non-specific...
 

well, database activity and file compression could still reflect many things. for instance, some hardware will throttle the CPU when it overheats, which could happen if a fan dies or a heatsink becomes dislodged. OTOH both symptoms you mention are IO-based, so perhaps a disk is sick (disks can get very slow as they're failing, for instance.) you've verified it's not something simpler, like an out-of-control daemon using all the CPU or memory?

Hi Mark

Thanks for the reply, I will go ahead and check the disk status, but noticeable point is that the same problem is in 4 servers.

You aren't saying how large your databases are. Also, how much MEM/CPU/IO do you have available?

Databases eat resources, and generally when query tuning and scheduler aren't options, then you need to throw more RAM, CPU, and faster disk.

Hi

Sorry for the late. Database size will be above 2 GB. but a fresh Linux box with no database also performing slow.

Seems kernel problem, the same kernel having pcs are performing slow but updating kernel also not worked.

I suggest you select one fast or faster performing server and a slow one. Collect hardware and configuration data. Install SAR or atsar, dstat or collectl and capture full stats for at least 48 hours. After comparing your next post then should then show the difference in systems hardware, configuration and performance-wise (or attach plain text files) because just saying "it's slow" without posting relevant data does not really help (most of us) help you in an efficient way.

It sounds like you have a bunch of servers, some of which perform well and others which don't. Perfect! Using the collectl-utility colmux, which I just released a couple of hours ago, you can monitor all the boxes as a set in real-time, sorting by memory usage, disk/network load, cpu or even process and/or nfs is that is part of the mix. Perhaps that will help identify what's different!

see - http://collectl-utils.sourceforge.net/colmux.html

-mark

Hi All,

My regret for not posting for a long time.

The problem is that the Hard disk drive which is installed in server is SATA but the disk detected in Linux OS was /dev/hda instead of /dev/sda. I change the settings of HDD to enchanced in BIOS. Now the HDD detects as /dev/sda and now it is performing faster as the other.