My testers pounded on my Alpha today and brought it to its knees.
We are running a redbrick warehouse with 4 CPUS _at_ 440MHz and 4 GB of RAM on
an AS8400 (DU4.0d patched).
They were definitely throwing a lot at the box, but I don't think
what they were doing should have had this impact:
vmstat:
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
32361 53 488K 48 25K 7M 268K 1M 46K 224K 14K 434 25K 2K
45 13 42
21371 58 488K 77 25K 1747 32 842 0 27 0 907 41K 4K
71 29 0
25368 58 488K 47 25K 1662 153 1135 23 175 0 1K 93K 6K
70 30 0
26368 57 488K 97 25K 2984 773 1110 428 505 0 1K 95K 10K
70 30 0
24370 57 488K 21 25K 1398 42 1234 445 44 0 1K 42K 4K
72 28 0
26369 57 488K 104 25K 1670 72 1468 74 51 0 1K 76K 5K
72 28 0
22373 57 488K 41 25K 2984 701 1316 26 442 0 1K 21K 4K
73 27 0
25375 56 488K 65 25K 775 27 704 29 41 0 1K 38K 5K
76 24 0
26375 58 488K 137 25K 4119 512 2743 66 228 0 1K 25K 6K
68 32 0
Swap:
> Total swap allocation:
> Allocated space: 1063830 pages (8311MB)
> Reserved space: 724723 pages ( 68%)
> In-use space: 164189 pages ( 15%)
> Available space: 339107 pages ( 31%)
>
>
Interesting that reserved space is so high but the in-use space is so low.
I've seen this before on other boxes, and don't quite understand what that's
all about. Anyone care to enlighten me?
I am also wondering about the huge number of shared message queues hanging
around. I'm beginning to suspect that memory is not getting cleaned out.
ipcs -q throws a lot of queues out, sometime they are days old for users no
longer on the system. Is there a relationship between messages queues and
swap space possibly?
Any clues or suggestions would be welcome. I'm trying to figure out if this
environment is broken somehow, before I spend a lot of money to beef up the
box.
TIA
susrod
Received on Sat Jul 10 1999 - 01:04:44 NZST