[Spread-users] spread delay problem

Adrian Revill adrian.revill at shazamteam.com
Tue Nov 4 12:12:41 EST 2008


We are seeing what looks like a pause in the spread communication of 2s.

The test set up is 2 master, and 3 worker nodes, the spread leader is 
one of the masters.
In the test, we apply load to one of the masters, each job is sent out 
to all the workers, the workers in this case just immediately send a 
reply back. This normally gives a round trip time for each job of 2-4ms
If the load is being applied to the master that is the spread leader, it 
works perfectly.
If the load is applied to master that is not the leader, we occasionally 
see 2s round trip times.
With spread running at normal priority it is approx 60 in 10000, and 
when we use a real time scheduler this drops to 6 in 10000

We assume this is due to the token being lost, but we can not find any 
sign of packet loss, as all the boxes are hardly loaded, running about 
4% CPU and that is mostly spreadd.

When we actually run in a mode that does some real work and on a larger 
cluster we see many more pauses

Can any one suggest anyway we can investigate further what is failing?

regards
Adrian




Just after start up

Monitor>
============================
Status at master2 V 4.00. 0 (state 1, gstate 1) after 301 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :     939      tok_hurry :     402     memb change:       2
sent pack:     192      recv pack :     869     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :     539      Aru       :     539     Highest seq:     539
Sessions :       4      Groups    :       9     Window     :      60
Deliver M:    1239      Deliver Pk:    1258     Pers Window:      15
Delta Mes:    1239      Delta Pack:     539     Delta sec  :     301
==================================

Monitor>
============================
Status at master1 V 4.00. 0 (state 1, gstate 1) after 305 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :     940      tok_hurry :     453     memb change:       2
sent pack:     266      recv pack :     795     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :     539      Aru       :     539     Highest seq:     539
Sessions :       4      Groups    :       9     Window     :      60
Deliver M:    1239      Deliver Pk:    1258     Pers Window:      15
Delta Mes:       0      Delta Pack:       0     Delta sec  :       4
==================================

Monitor>
============================
Status at opto-5 V 4.00. 0 (state 1, gstate 1) after 428 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :     939      tok_hurry :     471     memb change:       4
sent pack:     185      recv pack :     915     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :     539      Aru       :     539     Highest seq:     539
Sessions :       1      Groups    :       9     Window     :      60
Deliver M:    1277      Deliver Pk:    1308     Pers Window:      15
Delta Mes:      38      Delta Pack:       0     Delta sec  :     123
==================================

Monitor>
============================
Status at opto-7 V 4.00. 0 (state 1, gstate 1) after 423 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :     940      tok_hurry :     471     memb change:       4
sent pack:     184      recv pack :     916     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :     539      Aru       :     539     Highest seq:     539
Sessions :       1      Groups    :       9     Window     :      60
Deliver M:    1277      Deliver Pk:    1308     Pers Window:      15
Delta Mes:       0      Delta Pack:       0     Delta sec  :      -5
==================================

Monitor>
============================
Status at opto-6 V 4.00. 0 (state 1, gstate 1) after 426 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :     939      tok_hurry :     471     memb change:       4
sent pack:     184      recv pack :     916     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :     539      Aru       :     539     Highest seq:     539
Sessions :       1      Groups    :       9     Window     :      60
Deliver M:    1277      Deliver Pk:    1308     Pers Window:      15
Delta Mes:       0      Delta Pack:       0     Delta sec  :       3
==================================



After one test run

Monitor>
============================
Status at master2 V 4.00. 0 (state 1, gstate 1) after 655 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :   53272      tok_hurry :   10882     memb change:       2
sent pack:   50423      recv pack :   31746     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :   81790      Aru       :   81790     Highest seq:   81790
Sessions :       4      Groups    :       9     Window     :      60
Deliver M:   42490      Deliver Pk:   82509     Pers Window:      15
Delta Mes:   42490      Delta Pack:   81790     Delta sec  :     655
==================================

Monitor>
============================
Status at master1 V 4.00. 0 (state 1, gstate 1) after 660 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :   53273      tok_hurry :   11055     memb change:       2
sent pack:     614      recv pack :   81555     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :   81790      Aru       :   81790     Highest seq:   81790
Sessions :       4      Groups    :       9     Window     :      60
Deliver M:   42490      Deliver Pk:   82509     Pers Window:      15
Delta Mes:       0      Delta Pack:       0     Delta sec  :       5
==================================

Monitor>
============================
Status at opto-5 V 4.00. 0 (state 1, gstate 1) after 782 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :   53272      tok_hurry :   10951     memb change:       4
sent pack:   10360      recv pack :   71853     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :   81790      Aru       :   81790     Highest seq:   81790
Sessions :       1      Groups    :       9     Window     :      60
Deliver M:   42528      Deliver Pk:   82559     Pers Window:      15
Delta Mes:      38      Delta Pack:       0     Delta sec  :     122
==================================

Monitor>
============================
Status at opto-7 V 4.00. 0 (state 1, gstate 1) after 777 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :   53273      tok_hurry :   10951     memb change:       4
sent pack:   10363      recv pack :   71845     retrans    :       0
u retrans:       0      s retrans :       0     b retrans  :       0
My_aru   :   81790      Aru       :   81790     Highest seq:   81790
Sessions :       1      Groups    :       9     Window     :      60
Deliver M:   42528      Deliver Pk:   82559     Pers Window:      15
Delta Mes:       0      Delta Pack:       0     Delta sec  :      -5
==================================

Monitor>
============================
Status at opto-6 V 4.00. 0 (state 1, gstate 1) after 780 seconds :
Membership  :  5  procs in 1 segments, leader is master1
rounds   :   53272      tok_hurry :   10951     memb change:       4
sent pack:   10359      recv pack :   71849     retrans    :       5
u retrans:       5      s retrans :       0     b retrans  :       0
My_aru   :   81790      Aru       :   81790     Highest seq:   81790
Sessions :       1      Groups    :       9     Window     :      60
Deliver M:   42528      Deliver Pk:   82559     Pers Window:      15
Delta Mes:       0      Delta Pack:       0     Delta sec  :       3
==================================





______________________________________________________________________
This email has been scanned by the MessageLabs Email Security System.
For more information please visit http://www.messagelabs.com/email 
______________________________________________________________________




More information about the Spread-users mailing list