[Search for users]
[Overall Top Noters]
[List of all Conferences]
[Download this site]
Title: | Mailworks-unix |
Notice: | V2.0.4 now available -- see Note 4.37 5 |
Moderator: | TAMARA::NEUMAN::Neumann |
|
Created: | Wed Jun 02 1993 |
Last Modified: | Tue Jun 03 1997 |
Last Successful Update: | Fri Jun 06 1997 |
Number of topics: | 1384 |
Total number of notes: | 5851 |
1348.0. "System Freezes" by TAMARA::NEUMAN::Neumann (Stan Neumann) Wed Feb 12 1997 11:04
On a heavily loaded system, we have seen problems in which the
mail system effectively freezes for a few minutes (around
three minutes in the case we studied most closely) - after
this short freeze, the mail system will again start responding.
During the freeze, most other functions on the system continue
to work although some UNIX monitoring functions (such as
the top command, and ps auxw) will also freeze temporarily.
It appears that the operating system's virtual memory
manager has chosen to swap one of the MailWorks processes
out to disk, which effectively halts MailWorks until that
process can be swapped back in. Note that there is a
difference between paging, in which individual pages of
memory are moved between physical memory and disk, and
swapping, in which an entire process is written
out to disk. Paging is a normal part of a healthy
system; swapping takes place when the system feels that
it is working too hard at paging. Paging will not
stop the system (alhtough it can slow it down somewhat),
whereas swapping temporarily stops the application being
swapped.
When this problem occurs, it can be avoided by adjusting
the kernel tuning parameter vm-free-page-optimal.
In the file sysconfigtab, add the following entry
vm:
vm-free-page-optimal = 5
(note: if your sysconfigtab already has a vm: entry, place
the vm-free-page-optimal entry below it.)
After making this change, the operating system must be
rebooted for the change to take effect.
How can you tell if you are seeing this problem?
If you run a background process that consists only
of the command:
vmstat 10 > vmstat_logfile
then in the vmstat_logfile around the time of the
temporary freeze, you will see several things happening:
* Shortly before the freeze, free space will drop to a very
low level; usually the system idle time will also drop to
a low level.
* At the time of the freeze, the number of waiting threads
(the number in column 2 under "w") will drop suddenly,
then this number will rise suddenly when the freeze
is over.
* During the freeze, system idle time will usually rise
significantly.
* As a side effect of the freeze, you may see broken pipes,
network timeouts, and protocol errors in the MailWorks
log files.
If this problem occurs, it may indicate that the system needs
more physical memory. The tuning change above will avoid
the hard freeze of the system, but the system may still
be paging enough to affect the performance of the system.
-Stan
================================================================
Some background from the system tuning and performance manual:
If the number of pages in the free page list fall below the value
associated with the vm-free-page-optimal attribute for more than
five seconds, the task swapper (an extension of the page reclamation
code) is activated. The task swapper thread suspends processes,
writes to disk all of the dirtied pages associated with the
suspended processes, and places those pages on the free page list.
The task swapper first swaps out all swappable tasks that have been idle
for 30 seconds or more. If this does not satisfy the memory demand,
it begins swapping out the lowest priority tasks with the largest
resident set size, one at a time, until the memory demands are
satisifed (that is until the number of pages on the free page list
reaches the value associated with the vm-page-free-target
attribute.)
T.R | Title | User | Personal Name | Date | Lines
|
---|