[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: openldap 2.0.7 is incredibly slow

I would suggest you experiment with the Berkeley DB test
suite.  It might help isolate the problem (assuming you
are using Berkeley DB).


At 10:34 AM 1/10/01 +1100, James Gregory wrote:
>"Kurt D. Zeilenga" wrote:
>> Likely waiting for reverse lookups to timeout... you might
>> disable them via configure.
>Tried that and got exactly the same result. But this is fascinating. You
>see, out of desperation we decided to copy the entire hard disk of the
>production machine onto a spare IDE drive that was lying around the
>office. We then changed the lilo.conf file to point to the new IDE
>drive, and we changed it's fstab to mount the IDE drive. I would like to
>point out that this is /all/ we changed on the drive. We then booted
>this drive in my colleages Celeron 400 machine and the make test ran
>perfectly. We then put the IDE drive into the production machine, and
>disconnected the RAID controller, and amazingly enough it worked
>perfectly. Indeed, we put the raid controller back in the machine,
>booted off IDE and it still worked perfectly. Now, we then plugged the
>RAID array back in, so that it was in the original configuration. As
>expected it didn't work.
>So, to summarise, this is a problem which only appears when booting from
>the Compaq Linux Certified RAID 5 array.
>I've phoned compaq and they have said they'll get back to us twice now.
>If anyone has had any success or failure getting an even mildly similar
>configuration working, I would be thrilled to hear from you. We're
>getting desperate here.
>Thanks for all your help so far.
>> At 01:15 PM 1/5/01 +1100, James Gregory wrote:
>> >Hi everyone,
>> >
>> >I recenetly tried to move the openldap database from my test machine to
>> >the produciton machine for a project I'm working on. On my machine,
>> >(PIII-500) make test finishes reasonably quickly (say a couple of
>> >minutes). However, I've now built openldap 2.0.7 on the production
>> >machine and running make test takes several hours to complete. What's
>> >wrong? Several hours for a few little queries like that isn't fast
>> >enough for a production system.
>> >
>> >FYI, my machine is a pentium III-500, 64 MB ram, running mandrake 7.2.
>> >the production machine is a Pentium III-800 with 256 MB ram running
>> >redhat 6.2
>> >
>> >I'm using the same version of the berkeley ldbm back-end on both (rpms
>> >for my machine, built from source for the prod. machine). Same version
>> >of ssl (from the same rpms' on both machines) and the same version of
>> >cyrus-sasl on both machines (built from source on both machines)
>> >
>> >Apart from that everything on both machines is a pretty standard stock
>> >install of either distribution.
>> >
>> >oh, I also tried taking the compiled sources from my mandrake box and
>> >putting it on the production mahcine and running make test. It took
>> >hours.
>> >
>> >I'm guessing this a library issue and ldap is getting stuck trying to
>> >establish connections or something, because CPU utilisation is way down.
>> >The machine isn't doing any work at all.
>> >
>> >Has anyone run into this before?
>> >
>> >any ideas about how to solve it?
>> >
>> >thanks,
>> >
>> >james.