Bug 234801 - openldap got poor performance with RHEL5 on HP DL380 G4 and G5
Summary: openldap got poor performance with RHEL5 on HP DL380 G4 and G5
Status: CLOSED INSUFFICIENT_DATA
Alias: None
Product: Red Hat Enterprise Linux 5
Classification: Red Hat
Component: openldap   
(Show other bugs)
Version: 5.0
Hardware: i686
OS: Linux
medium
high
Target Milestone: ---
: ---
Assignee: Jan Zeleny
QA Contact:
URL:
Whiteboard:
Keywords:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2007-04-02 07:09 UTC by Ching-Che Yen
Modified: 2009-10-19 14:24 UTC (History)
3 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Environment:
Last Closed: 2009-10-19 14:24:05 UTC
Type: ---
Regression: ---
Mount Type: ---
Documentation: ---
CRM:
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---


Attachments (Terms of Use)

Description Ching-Che Yen 2007-04-02 07:09:03 UTC
Description of problem:

I created a ldap server(openldap-2.3.27-5) with RHEL5,I found the performance is
very bad on a very powerful machine(HP DL380 G5,2 way 3.2GHZ cpu,4GB Memory,SAS
disks),even I had tuned the slapd.conf and DB_CONFIG.I test it with my own
script , I found the ldap server could just handle 4 authentication queries per
second !!!

Then I tried to create ldap server  with RHEL5 on my Desktop computer(AMD
1.7Gz,2GB Memory,IDE disk),and I get greater ldap query performance on it (about
160 authentication queries per second !!!)

In order to test it further,I install RHEL4(openldap-2.2.13-6.4) and
RHEL3(openldap-2.0.27-22) on the HP DL380.All performance is very bad (about 4
to 12 authenticaion queries per second).

Then I guess that maybe it's the SMP or Hyper-threading to influence the
performance.
So I tried to unplug one cpu to reduce the number of cpu from 2 ways to 1
way,and disable Hyper-threading in the BIOS.
The result is.....still very bad.....

btw,I install the RHEL5 on another pc(p4 2.8Ghz 1GB ram),It works
smoothly.(about 140 authentication queries per second).

No I really don't know how to test it further....@_@
  
Could any one give me a suggesstion?

Thank you very much,and sorry for my poor english.

Comment 1 Jan Zeleny 2009-09-01 13:30:47 UTC
It has been over two years since this bugzilla was created. Is this issue still present or was there any progress?

Comment 2 Timm Stamer 2009-09-01 13:33:31 UTC
I've got a similar problem on a brand-new FSC RX200 S5 (Intel Nehalem, 12G RAM) running RHEL 5.3 x86_64.

In the first minute after enabling this host on the upstream loadbalancer (Foundry ServerIron GT E2) everything working fine. LDAP answers takes ~0.1 seconds. After some time connections get stucked and answers take up to 5 seconds or got timed out.

Other servers in the cluster are much older xeon processors w 2gb of RAM running RHEL AS4 (i386). One of these is ldap master, the new one is a replica.


regards,
timm

Comment 3 Timm Stamer 2009-09-01 13:35:14 UTC
sorry jan, several "human interrupts" delayed my post ;)

Comment 4 Jan Zeleny 2009-09-01 15:23:45 UTC
Thank you for the info. I suggest consulting these issues with customer support, those guys might have some tips about it and there might even be no need for any patch.

Comment 5 Ching-Che Yen 2009-09-03 06:35:54 UTC
This problem was solved after about 2 months.

I think that the yum update some rpm packages to fix it.

Then I checked the yum.log file,but there are no openldap or some similar updates.

I had tried to update the rpm packages in the yum.log one bye one manually to try to repeat the problem,but I just can't repeat it anymore.(to many packages and dependence problems)

Now I just know that if I install the newest RHEL,the openldap server is ok now.

Comment 6 Jan Zeleny 2009-10-19 14:24:05 UTC
Original reporter of this bug solved the issue and I have no information to consider this to be a bug, so I'm closing it. If anyone else has similar troubles, please contact our customer support, they might help you with debugging before it gets to me as a new bug.


Note You need to log in before you can comment on or make changes to this bug.