[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
Re: 8 hours tests ends with inconsistent DB.
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Trevor Warren wrote:
| Hello Quanah,
|
| I totally agree with you on the same. Infact this is
| the norm for all the performance tests carried out by
| us for any system under test.
|
| We started out with 1 Lakh records and are still at
| one lakh since the god damn thing manages to go down
| at the drop of a hat.
|
| I just ended an 8 hour test on my HP DL580 G1
| 4cpu-4Gb machine for 200 users with think time.
| Writes/s are 28 but the server aborted with the same
| malloc error as in yesterdays overnite test.
|
| Qunanh.....all said and done unless this setup can
| scale atleast for 1 Lakh users i can't recommend it
| for my setups across 50 countries across the world.
| More importantly i am very interested ni seeing Open
| Source scale.
Well, to see it "scale", you need to start at the bottom and work up.
Otherwise you have just tried a large-scale deployment (and done no
scaling).
|
| I humbly hope my feedback is helping developers
| understand issues and scale the app by elimnation of
| the code bottlenecks as soon as possible.
|
| BTW: My aim was to get to 50million users...:D. Its
| one of the largest systems worldwide we would have
| ever built if we ever did.
Well, I have seen absolultely no problems like this on automated testing
with batch imports of 250000 entries (approx 80 minutes via ldapadd with
transaction logging enabled and with 2 slaves), followed query sets by
12-16 simultaneous clients while the LDAP service was being migrated
between node members (serving about 11000 reads/minute).
I was running on Dell 1650s (P3-1GHz/1GB ram) writing to an EMC
Clariion, running Openldap-2.1.{25,29,30} on db-4.2.52.2 on Red Hat
Enterprise 2.1.
I would suggest you start at a similar position (with a "stable" release
of openldap with a reasonable size database), and scale from there.
However, I simply can't agree with your statement that the "damn thing
manages to go down at the drop of a hat". I haven't seen one instance of
a slapd crash in my tests (but, I am not doing any SASL/krb5 stuff for
instance).
Regards,
Buchan
- --
Buchan Milne Senior Support Technician
Obsidian Systems http://www.obsidian.co.za
B.Eng RHCE (803004789010797)
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org
iD8DBQFAyb9srJK6UGDSBKcRAn13AKCNrwam6LttFb0urs1bVgjFpbpTAgCeLLTz
Gu2BDD4YoTfaThB+j2gTTkQ=
=Df7C
-----END PGP SIGNATURE-----