Problem was identified in AIX 4.3.3 with async i/o which caused Oracle systems to crash. Patch was applied to AIX which cured the Oracle problem but pushed my batch run time by an extra hour. On running vmstats noticed that the sytem figure was consistanly running between 30 and 40% with user running about 20% below that. This cannot be right and still think that there is problem with asyn i/o with AIX, anybody else seen this problem or any comments on the high system figures?
Tek-Tips is the largest IT community on the Internet today!
Members share and learn making Tek-Tips Forums the best source of peer-reviewed technical information on the Internet!
-
Congratulations IamaSherpa on being selected by the Tek-Tips community for having the most helpful posts in the forums last week. Way to Go!
AIX 4.3.3 and Async i/o problems with Oracle 8.1.6
- Thread starter davemak
- Start date
- Thread starter
- #3
Our test plan is to switch async i/o off in Oracle first and monitor then if no changes switch off in AIX, however this does not fix the problem. Have also just got the latest maint disk (8) for AIX which I will be applying. Our AIX support company cannot understand why the 'system' figure on vmstats is high. One thought, our paging is not high usually round 2 or 3 but I have noticed that our free pages does drop below 120 a few times overnight I know this triggers the system to free pages but would it drive the system figure that high.
Sample vmstats:
The first column is a time stamp we have added the rest of the columns are as per the vmstat commnad.
01:10:39 2 6 88748 276 0 0 1 255 585 0 1144 26383 2795 23 15 28 34
01:15:39 2 6 89878 259 0 1 1 188 515 0 1189 31551 2634 24 16 26 33
01:20:40 1 7 91062 150 0 3 4 305 921 0 1331 30689 1485 14 15 27 44
01:25:40 1 6 89197 508 0 0 1 94 253 0 1226 19101 1482 10 10 39 41
01:30:40 1 6 88729 742 0 0 1 100 293 0 1172 4669 1493 9 8 44 38
01:35:41 1 6 92960 121 0 1 1 141 355 0 1162 3906 1269 9 9 42 40
01:40:41 1 5 92300 122 0 1 4 225 753 0 1109 3069 1098 10 8 47 35
01:45:41 1 6 91861 124 0 2 0 251 641 0 1135 2634 1071 10 8 43 40
01:50:41 1 6 91002 122 0 3 1 209 571 0 1111 3773 1231 11 8 43 38
01:55:42 1 5 90077 122 0 2 0 117 379 0 1095 11249 1071 6 11 47 36
02:00:42 2 4 89579 512 0 0 0 9 30 0 964 79019 1094 7 30 44 19
02:05:42 2 4 89620 387 0 0 0 7 28 0 928 74977 1052 7 30 47 17
02:10:43 1 4 89144 991 0 0 0 0 0 0 904 68491 978 5 21 58 16
Sample vmstats:
The first column is a time stamp we have added the rest of the columns are as per the vmstat commnad.
01:10:39 2 6 88748 276 0 0 1 255 585 0 1144 26383 2795 23 15 28 34
01:15:39 2 6 89878 259 0 1 1 188 515 0 1189 31551 2634 24 16 26 33
01:20:40 1 7 91062 150 0 3 4 305 921 0 1331 30689 1485 14 15 27 44
01:25:40 1 6 89197 508 0 0 1 94 253 0 1226 19101 1482 10 10 39 41
01:30:40 1 6 88729 742 0 0 1 100 293 0 1172 4669 1493 9 8 44 38
01:35:41 1 6 92960 121 0 1 1 141 355 0 1162 3906 1269 9 9 42 40
01:40:41 1 5 92300 122 0 1 4 225 753 0 1109 3069 1098 10 8 47 35
01:45:41 1 6 91861 124 0 2 0 251 641 0 1135 2634 1071 10 8 43 40
01:50:41 1 6 91002 122 0 3 1 209 571 0 1111 3773 1231 11 8 43 38
01:55:42 1 5 90077 122 0 2 0 117 379 0 1095 11249 1071 6 11 47 36
02:00:42 2 4 89579 512 0 0 0 9 30 0 964 79019 1094 7 30 44 19
02:05:42 2 4 89620 387 0 0 0 7 28 0 928 74977 1052 7 30 47 17
02:10:43 1 4 89144 991 0 0 0 0 0 0 904 68491 978 5 21 58 16
Dave,
You're doing all of the things that I would be doing. (if that's any consolation <wry smile>)
Applying the latest fixes to AiX and then re-doing your tests is certainly a good idea, time consuming though :-(
In your position I would be putting a fair amount of pressure on my AiX support company to escalate the problem up to IBM and on Oracle Support as well, to explain why their kernel fix had such an effect upon performance.
It *may* be that you'll have to live with the performance hit and attempt to improve things using other methods... Mike
michael.j.lacey@ntlworld.com
Email welcome if you're in a hurry or something -- but post in tek-tips as well please, and I will post my reply here as well.
You're doing all of the things that I would be doing. (if that's any consolation <wry smile>)
Applying the latest fixes to AiX and then re-doing your tests is certainly a good idea, time consuming though :-(
In your position I would be putting a fair amount of pressure on my AiX support company to escalate the problem up to IBM and on Oracle Support as well, to explain why their kernel fix had such an effect upon performance.
It *may* be that you'll have to live with the performance hit and attempt to improve things using other methods... Mike
michael.j.lacey@ntlworld.com
Email welcome if you're in a hurry or something -- but post in tek-tips as well please, and I will post my reply here as well.
- Status
Similar threads
- Locked
- Question