Problem with Swap on Sun Ultra running Solaris 2.5.1
I noticed (today) that the the allocation for swap is changing (not fixed) overtime, and causing the system to crash (/tmp full). Everytime, I had to recycle the server.
The mahicne is Sun Ultra running Solaris 2.5.1.
The questions are: What is causing this? Is this normal. And the best part is, how to fix it.
At the moment, I had to do a quick fix, creating another swapfile round 256MB and add it to swap, so that a have longer time between recycling (sigh..).
Some info. (more if required)).
Disk Layout.
Part Tag Flag Cylinders Size Blocks
0 root wm 0 - 242 180.35MB (243/0/0) 369360
1 unassigned wm 0 0 (0/0/0) 0
2 backup wm 0 - 2732 1.98GB (2733/0/0) 4154160
3 unassigned wm 0 0 (0/0/0) 0
4 usr wm 243 - 2386 1.55GB (2144/0/0) 3258880
5 unassigned wm 0 0 (0/0/0) 0
6 swap wu 2387 - 2731 256.05MB (345/0/0) 524400
7 unassigned wm 0 0 (0/0/0) 0
# showrev
Hostname: myserver
Hostid: somehostid
Release: 5.5.1
Kernel architecture: sun4u
Application architecture: sparc
Hardware provider: Sun_Microsystems
Domain: domain.com
Kernel version: SunOS 5.5.1 Generic 103640-27 March 1999
/etc/vfstab
swap - /tmp tmpfs - yes -
/opt/swap/swap1 - - swap - no -
# swap -l
swapfile dev swaplo blocks free
/dev/dsk/c0t0d0s6 32,6 16 524384 524384
/opt/swap/swap1 - 16 524272 524272
Swapfile reading over time (10 seconds interval).
Filesystem kbytes used avail capacity Mounted on
swap 695144 8 695136 1% /tmp
swap 695096 8 695088 1% /tmp
swap 695776 8 695768 1% /tmp
swap 691688 8 691680 1% /tmp
swap 693968 8 693960 1% /tmp
swap 696552 8 696544 1% /tmp
swap 699008 8 699000 1% /tmp
swap 698080 8 698072 1% /tmp
swap 700496 8 700488 1% /tmp
swap 698240 8 698232 1% /tmp
swap 699840 8 699832 1% /tmp
swap 698376 8 698368 1% /tmp
swap 701208 8 701200 1% /tmp
swap 701288 8 701280 1% /tmp
swap 700264 8 700256 1% /tmp
swap 698704 8 698696 1% /tmp
swap 697184 8 697176 1% /tmp
swap 695184 8 695176 1% /tmp
swap 686656 8 686648 1% /tmp
I would appreciate any help.
Samri.
The mahicne is Sun Ultra running Solaris 2.5.1.
The questions are: What is causing this? Is this normal. And the best part is, how to fix it.
At the moment, I had to do a quick fix, creating another swapfile round 256MB and add it to swap, so that a have longer time between recycling (sigh..).
Some info. (more if required)).
Disk Layout.
Part Tag Flag Cylinders Size Blocks
0 root wm 0 - 242 180.35MB (243/0/0) 369360
1 unassigned wm 0 0 (0/0/0) 0
2 backup wm 0 - 2732 1.98GB (2733/0/0) 4154160
3 unassigned wm 0 0 (0/0/0) 0
4 usr wm 243 - 2386 1.55GB (2144/0/0) 3258880
5 unassigned wm 0 0 (0/0/0) 0
6 swap wu 2387 - 2731 256.05MB (345/0/0) 524400
7 unassigned wm 0 0 (0/0/0) 0
# showrev
Hostname: myserver
Hostid: somehostid
Release: 5.5.1
Kernel architecture: sun4u
Application architecture: sparc
Hardware provider: Sun_Microsystems
Domain: domain.com
Kernel version: SunOS 5.5.1 Generic 103640-27 March 1999
/etc/vfstab
swap - /tmp tmpfs - yes -
/opt/swap/swap1 - - swap - no -
# swap -l
swapfile dev swaplo blocks free
/dev/dsk/c0t0d0s6 32,6 16 524384 524384
/opt/swap/swap1 - 16 524272 524272
Swapfile reading over time (10 seconds interval).
Filesystem kbytes used avail capacity Mounted on
swap 695144 8 695136 1% /tmp
swap 695096 8 695088 1% /tmp
swap 695776 8 695768 1% /tmp
swap 691688 8 691680 1% /tmp
swap 693968 8 693960 1% /tmp
swap 696552 8 696544 1% /tmp
swap 699008 8 699000 1% /tmp
swap 698080 8 698072 1% /tmp
swap 700496 8 700488 1% /tmp
swap 698240 8 698232 1% /tmp
swap 699840 8 699832 1% /tmp
swap 698376 8 698368 1% /tmp
swap 701208 8 701200 1% /tmp
swap 701288 8 701280 1% /tmp
swap 700264 8 700256 1% /tmp
swap 698704 8 698696 1% /tmp
swap 697184 8 697176 1% /tmp
swap 695184 8 695176 1% /tmp
swap 686656 8 686648 1% /tmp
I would appreciate any help.
Samri.
Last Comment
ASKER CERTIFIED SOLUTION
THIS SOLUTION IS ONLY AVAILABLE TO MEMBERS.
View this solution by signing up for a free trial.
Members can start a 7-Day free trial and enjoy unlimited access to the platform.
But it can also be a file filling up /tmp which is written by a process, so also log file sizes within /tmp (and subdirs).
ASKER
Thanks all,
Basicall everybody's coment are good. (How do I give points :?)
Anyway, I'm wandering if there is any site that I can visit to read more detail on this (perhaps any TID on Sun website).
And looking at my quick-fix (adding another swapfile) - is it the right way to get around, or anybody has a better recommendation.
thanks again. (i'll keep the question open for a few more days)..
samri.
Basicall everybody's coment are good. (How do I give points :?)
Anyway, I'm wandering if there is any site that I can visit to read more detail on this (perhaps any TID on Sun website).
And looking at my quick-fix (adding another swapfile) - is it the right way to get around, or anybody has a better recommendation.
thanks again. (i'll keep the question open for a few more days)..
samri.
The amount of swap space that you need depends on how much ram you have, and how much work you are trying to get done. A general rule of thumb is to make swap = 2 X ram, but that is just a guess. Solaris mounts /tmp on swap, and this is both a good thing and also a bad thing.
As mentioned previously, the amount of available swap at any given time is egual to (total ram + total swap) - (total memory used by all processes + total size of all files stored on /tmp).
Having /tmp mounted on swap is good because small temporary files will reside in memory on a lightly loaded machine, and accessing them will be very quick. Having /tmp on swap is bad because uncompressing that 300MB webserver logfile in there chews up swap space that could be used to start new processes.
Sometimes, in order to keep the box running while you try to discover what is filling swap up, allocating additional swap file(s) as you did will help see you through.
One way to prevent large files in /tmp from consuming all of your swap space is to create a disk partition and mount /tmp on it. In that case if /tmp fills up, you will still have swap space available to start new processes.
You didn't mention what type of work this machine is doing, or how much ram is installed.
Good luck, hope this helps,
--frankf
As mentioned previously, the amount of available swap at any given time is egual to (total ram + total swap) - (total memory used by all processes + total size of all files stored on /tmp).
Having /tmp mounted on swap is good because small temporary files will reside in memory on a lightly loaded machine, and accessing them will be very quick. Having /tmp on swap is bad because uncompressing that 300MB webserver logfile in there chews up swap space that could be used to start new processes.
Sometimes, in order to keep the box running while you try to discover what is filling swap up, allocating additional swap file(s) as you did will help see you through.
One way to prevent large files in /tmp from consuming all of your swap space is to create a disk partition and mount /tmp on it. In that case if /tmp fills up, you will still have swap space available to start new processes.
You didn't mention what type of work this machine is doing, or how much ram is installed.
Good luck, hope this helps,
--frankf
ASKER
Blowfish,
Thanks for the elaborated explanation, some which I already knew, but most of them are quite informative I would say.
As for you curiosity, the machine is a mailhub for a corporate network (16000+ users). It's running Sendmail 8.9.3. Haas apache, and mrtg running for internal monitoring. And nothing else (except for the usual stuff that comes with Solaris 2.5.1).
The machine has 128MB Physical RAM, 256MB Swap partition (c0t3d0s6). Single CPU, UltraSparc 167MHZ (i think). We are on a 2Mbit link to our ISP.
Initially it runs fine (for more than 2 years already), except it went down a couple of days ago. Reason: /tmp full. Quick fix, I added another 256MB swapfile -- and it looks okay for the last few days.
Any hints on performance improvement.
BTW, how do I spread the pts across the experts that contributed the quality comments (see Customer Service.. huh :)
Thanks again to all.
samri
Thanks for the elaborated explanation, some which I already knew, but most of them are quite informative I would say.
As for you curiosity, the machine is a mailhub for a corporate network (16000+ users). It's running Sendmail 8.9.3. Haas apache, and mrtg running for internal monitoring. And nothing else (except for the usual stuff that comes with Solaris 2.5.1).
The machine has 128MB Physical RAM, 256MB Swap partition (c0t3d0s6). Single CPU, UltraSparc 167MHZ (i think). We are on a 2Mbit link to our ISP.
Initially it runs fine (for more than 2 years already), except it went down a couple of days ago. Reason: /tmp full. Quick fix, I added another 256MB swapfile -- and it looks okay for the last few days.
Any hints on performance improvement.
BTW, how do I spread the pts across the experts that contributed the quality comments (see Customer Service.. huh :)
Thanks again to all.
samri
Community Support has reduced points from 100 to 50
Hello everyone,
Reducing points to one half to allow for split.
samri, you can now accept one of the comments in this thread as an answer. To award the other Expert, you can create a new question in this topic area with a title of 'For ExpertName -- 10441857' using that Experts username.
Remember, the Accept Comment as Answer button is in the header of the comment.
For your convenience, you can use this link to create the new question:
https://www.experts-exchange.com/bin/NewQForm?ta=3316
darinw
Customer Service
Reducing points to one half to allow for split.
samri, you can now accept one of the comments in this thread as an answer. To award the other Expert, you can create a new question in this topic area with a title of 'For ExpertName -- 10441857' using that Experts username.
Remember, the Accept Comment as Answer button is in the header of the comment.
For your convenience, you can use this link to create the new question:
https://www.experts-exchange.com/bin/NewQForm?ta=3316
darinw
Customer Service
ASKER
Comment accepted as answer
ASKER
thanks to all.
Unix OS
Unix is a multitasking, multi-user computer operating system originally developed in 1969 at Bell Labs. Today, it is a modern OS with many commercial flavors and licensees, including FreeBSD, Hewlett-Packard’s UX, IBM AIX and Apple Mac OS-X. Apart from its command-line interface, most UNIX variations support the standardized X Window System for GUIs, with the exception of the Mac OS, which uses a proprietary system.
33K
Questions
--
Followers
--
Top Experts
Get a personalized solution from industry experts
TRUSTED BY
/usr/ucb/ps -aux
periodically and redirect the output to a file. This should help you identify which processes are using up the memory. Look at the %MEM and SZ fields.
Hope this is useful.