6.0.0-beta1
10/18/25

[#5728] horde3_prefs table in IMP Database very poor performance
Summary horde3_prefs table in IMP Database very poor performance
Queue Horde Base
Queue Version 3.1.3
Type Bug
State Duplicate
Priority 2. Medium
Owners
Requester lbellamy (at) deakin (dot) edu (dot) au
Created 09/18/2007 (6605 days ago)
Due
Updated 09/18/2007 (6605 days ago)
Assigned
Resolved 09/18/2007 (6605 days ago)
Github Issue Link
Github Pull Request
Milestone
Patch No

History
09/18/2007 03:28:22 AM Michael Rubinsky Comment #3
State ⇒ Duplicate
Reply to this comment
Duplicate of Bug: 4079




09/18/2007 03:20:58 AM lbellamy (at) deakin (dot) edu (dot) au Comment #2 Reply to this comment
Just to add to this; would it be advisable to document the use of 
indexing on certain tables in the imp database to prevent other users 
from experiencing this issue?



Regards,



Luke
09/18/2007 03:19:18 AM lbellamy (at) deakin (dot) edu (dot) au Comment #1
State ⇒ Unconfirmed
Priority ⇒ 2. Medium
Type ⇒ Bug
Summary ⇒ horde3_prefs table in IMP Database very poor performance
Queue ⇒ Horde Base
Reply to this comment
*informational only; we have mitigated this bug from our side*



We currently have a horde 3.1.3 install running IMP as our university 
Web Mail client. On the same host we run webmail from, we have the imp 
database. We have a potential userbase of this application of 40,000 
users.



Recently we noticed that we had rather high load on this host  (in 
excess of 100....) in comparison to the 2 or 3 it normally sits on.



Investigation lead us to notice that imp was generating some 
inefficient queries to the imp database, and more specifically to the 
horde3_prefs table. In this table, we have approx 350,000 rows, and 
the only entry into this table was via a full table scan. During the 
times of high load, imp was causing our oracle backend to do about 6.6 
billion fetched records; understandably causing high load.



We resolved this by placing indexing on the horde3_prefs table so full 
table scans are not required every single query, which has 
significantly reduced the load placed on the server by horde/imp/oracle.



This is just to let you know that it would appear once we hit a 
certain user / preference threshold, performance of the application 
was significantly impacted, to the point where it caused issues with 
the server it was running on.



If you need any more information, please dont hesitate to send me an email.



Regards,



Luke

Saved Queries