Sincere thanks to Olle Eriksson for the lightning response
that put me out of my misery.
[snip]
From: Olle Eriksson <olle_at_cb.uu.se>
First of all you should delete the link /sbin/swapdefault. That enables
lazy mode which means that swap space is claimed when needed and not
reserverd when the program starts.
[/snip]
Now I have all my swap space free with all webservers
and processes running.
Swap partition /dev/rz0b:
Allocated space: 16384 pages (128MB)
In-use space: 63 pages ( 0%)
Free space: 16321 pages ( 99%)
Total swap allocation:
Allocated space: 16384 pages (128MB)
In-use space: 63 pages ( 0%)
Available space: 16321 pages ( 99%)
I'd been looking for /etc/swapdefault then spent the
morning looking thru the ORA sys admin book.
Happy camper again. Thank you again,
chas
Original Post :
---------------
>Running into continuous "swap space below 10 percent free"
>warning messages when Netscape Enterprise 3.5 is initialised.
>
>-------- Before any webservers are started : ----------
>Swap partition /dev/rz0b (default swap):
> Allocated space: 16384 pages (128MB)
> In-use space: 40 pages ( 0%)
> Free space: 16344 pages ( 99%)
>
>Total swap allocation:
> Allocated space: 16384 pages (128MB)
> Reserved space: 5143 pages ( 31%)
> In-use space: 40 pages ( 0%)
> Available space: 11241 pages ( 68%)
>
>Virtual Memory Statistics: (pagesize = 8192)
> procs memory pages intr cpu
> r w u act free wire fault cow zero react pin pout in sy cs us sy
> id
> 3 35 16 5124 5167 1080 25K 6998 6130 89 5275 0 21 246 68 3 4
> 93
> 2 36 16 5252 5030 1089 338 168 53 0 23 0 7 256 62 3 3
> 95
> 2 36 16 5252 5030 1089 29 0 28 0 0 0 0 190 46 0 1
> 99
> 2 36 16 5252 5030 1089 28 0 28 0 0 0 1 191 49 0 1
> 99
>
>
>
>-------- With one instance of the webserver running : -----
>
># swapon -s
>Swap partition /dev/rz0b (default swap):
> Allocated space: 16384 pages (128MB)
> In-use space: 389 pages ( 2%)
> Free space: 15995 pages ( 97%)
>
>
>Total swap allocation:
> Allocated space: 16384 pages (128MB)
> Reserved space: 10703 pages ( 65%)
> In-use space: 389 pages ( 2%)
> Available space: 5681 pages ( 34%)
>
>Virtual Memory Statistics: (pagesize = 8192)
> procs memory pages intr cpu
> r w u act free wire fault cow zero react pin pout in sy cs us sy
> id
> 4 49 16 7526 2621 1225 30K 7829 8070 154 6914 0 26 303 91 3 4
> 93
> 2 51 16 7542 2599 1231 332 76 58 0 22 0 0 305 170 2 2
> 95
> 4 49 16 7542 2599 1231 35 0 34 0 0 0 1 272 174 0 1
> 99
> 3 50 16 7555 2586 1231 76 7 40 0 0 0 27 264 185 2 1
> 97
>
>
>--------- With the second instance started : -----------
>
># swapon -s
>Swap partition /dev/rz0b (default swap):
> Allocated space: 16384 pages (128MB)
> In-use space: 389 pages ( 2%)
> Free space: 15995 pages ( 97%)
>
>Total swap allocation:
> Allocated space: 16384 pages (128MB)
> Reserved space: 15362 pages ( 93%)
> In-use space: 389 pages ( 2%)
> Available space: 1022 pages ( 6%)
>
>Virtual Memory Statistics: (pagesize = 8192)
> procs memory pages intr cpu
> r w u act free wire fault cow zero react pin pout in sy cs us sy
> id
> 5 56 16 9249 816 1308 36K 8399 10K 350 8963 14 31 354 115 4 4
> 91
> 3 58 16 9265 794 1314 348 78 62 28 22 0 0 406 216 3 3
> 94
> 2 59 16 9264 795 1314 38 0 36 0 0 0 0 293 218 0 2
> 98
> 2 59 16 9264 795 1314 139 29 36 0 2 0 9 295 231 1 1
> 98
>
>
>Refering to the Digital webserver tuning paper (found
>at http://www.digital.com/internet/document/ias/tuning.html),
>I used the dxkerneltuner to change the vm's vm_mapentries
>from 200 to 20000 and rebooted.
>This didn't help much. (in fact the above vmstat is taken with
>vm_mapentries at 20000)
>
>Past summaries in the archive suggested that increasing the
>vm_mapentries did the trick; others suggested deleting or
>renaming the symlink /etc/swapdefault but that doesn't even
>exist on my box.
>
>So, I figured I could configure the process (in this case, Netscape
>Ent. server 3.5) to reserve or use less memory.
>But the only 'performance tuning' available is to reduce the
>Max No of Simultaneous Requests from 512 which I don't really
>want to do. (and I've messed up my magnus.conf
>and other config files enough to know not to edit them by hand anymore).
>
>Unfortunately I have no free partitions to allocate
>another swap space, and in any case this doesn't seem like
>a scalable solution judging by the amount of swapspace each
>instance of netscape reserves. I should be able to hang 10
>webservers off this machine (at least) but not at the current
>rate (about 36 MB swap space per instance).
>
>Any advice much appreciated,
>
>thank you very much,
>
Received on Tue May 05 1998 - 08:42:31 NZST