[rt-users] rt 3.4.4 - too slow in some situations
jesse at bestpractical.com
Sun Dec 4 14:07:23 EST 2005
On Sun, Dec 04, 2005 at 08:22:40PM +0200, Victor wrote:
> It seems the postgresql was the bottlenec.
> Even if it doesn't takes to mutch from the processor, the cost is to high.
> Cost is the time spend accesing the hdd.
Please try out RT 3.4.5rc2 to see if it fixes this for you.
> This is where it stays too much:
> EXPLAIN SELECT DISTINCT main.* FROM Users main , Principals
> Principals_1, ACL ACL_2, Groups Groups_3, CachedGroupMembers
> CachedGroupMembers_4 WHERE ((ACL_2.RightName = 'OwnTicket')) AND
> ((CachedGroupMembers_4.MemberId = Principals_1.id)) AND ((Groups_3.id =
> CachedGroupMembers_4.GroupId)) AND ((Principals_1.Disabled =
> '0')or(Principals_1.Disabled = '0')) AND ((Principals_1.id != '1')) AND
> ((main.id = Principals_1.id)) AND ( ( ACL_2.PrincipalId =
> Groups_3.id AND ACL_2.PrincipalType = 'Group' AND ( Groups_3.Domain =
> 'SystemInternal' OR Groups_3.Domain = 'UserDefined' OR Groups_3.Domain =
> 'ACLEquivalence')) OR ( ( (Groups_3.Domain = 'RT::Queue-Role' ) ) AND
> Groups_3.Type = ACL_2.PrincipalType) ) AND (ACL_2.ObjectType =
> 'RT::System' OR (ACL_2.ObjectType = 'RT::Queue') ) ORDER BY
> main.Name ASC;
> And the total cost is:
> Unique (cost=5057.11..5057.20 rows=1 width=1135)
> I remaind at the solution with the text instead of the select box for
> the "value of actor".
> Even if this select would work better, mozilla takes o lot to print this
> page because is big.
That sounds like you've mistakenly granted "Everybody" the right to
> Jesse Vincent wrote:
> >On Sat, Dec 03, 2005 at 02:10:34PM +0200, Victor wrote:
> >>By example it takes 35 secons to display the tickets->New query page.
> >>I have a big database in Pg 8.1.
> >>I also tried with diffrent versions of mysql, but the best result was
> >>with postgres.
> >>I have a very good computer(2Gb ram, CPU 2x20Gh each multithread, 512kb
> >>When I try to load the page tickets->New, apache process ocupies 100% of
> >>one processor for about 35 sec.
> >What do postgres' logs say about what's taking time?
More information about the RT-Users