The highest priority swap should always be VDISK (not real disk).
Save the real disk for a small contingency (with the lowest priority) in
case you run out of both MEM and VDISK swap.

Gaylord Toneff
IBM Global Services
gton...@us.ibm.com

Member, z/Linux Commercial Account Support
Kaiser Permanente Account
Home Office  661 456 2242
Cell 661 618 2825
Kaiser Office 626 564 7473


                                                                       
  From:       "Bauer, Bobby (NIH/CIT) [E]" <baue...@mail.nih.gov>      
                                                                       
  To:         LINUX-390@vm.marist.edu                                  
                                                                       
  Date:       03/03/2011 04:48 AM                                      
                                                                       
  Subject:    Spiking server                                           
                                                                       
  Sent by:    Linux on 390 Port <LINUX-390@vm.marist.edu>              
                                                                       





We have a Wordpress server that really spikes during certain, know times of
the month, about 45000 hits/hour. Its running on a single z9 IFL with only
4G of memory on the lpar, z/VM 5.4, REHL 6 (yeah, I know, more memory, good
luck since we are a govt. agency). The user did not expect this kind of
response so we have all been surprised.

We have Supercache in use.

At 1G of memory it crashed yesterday due to lack of memory so we upped it
to 2G and are waiting for the next cycle. It has swap space of:
swapon -s
Filename                                Type            Size    Used
Priority
/dev/dasda2                             partition       1023976 0       -1
/dev/dasdb1                             partition       194964  0       2
/dev/dasdc1                             partition       64976   0       1

dasda2 is real dasd
dasdb1 and c1 are VDISK defined using the swapgen macro from Sine Nomine

this morning the server looks like this:
free
             total       used       free     shared    buffers     cached
Mem:       2050360    1323728     726632          0     114588     345964
-/+ buffers/cache:     863176    1187184
Swap:      1283916          0    1283916


So the general question is, are there other steps we can take to help
response time when usage peaks? There are 2 other production servers on
this lpar. One is very low usage, the other has the potential for the same
kind of activity. There is a test lpar sharing the IFL with 4G of memory
also. I've thought of stealing a G from test and moving it to production.
There are plans to host Wiki's on the production lpar also.
Any suggestions would be appreciated.

Bobby Bauer
Center for Information Technology
National Institutes of Health
Bethesda, MD 20892-5628
301-594-7474



----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or
visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/


----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390
----------------------------------------------------------------------
For more information on Linux on System z, visit
http://wiki.linuxvm.org/

<<inline: graycol.gif>>

<<inline: ecblank.gif>>

Reply via email to