The mailing list for listmasters using Sympa

[Steve Shipway <address@concealed>]

Wondering if anyone here has experienced the same issues as us.

 

According to the documentation, the bulk_lazytime setting in the Sympa.conf represents the number of seconds that a bulk thread can have no work before it will exit.

 

On our system (6.1.23) it seems that, while the various other bulk_* options are working to gradually increase the number of bulk threads, once started they never die.  Thus, we quickly hit our maximum number of bulk processes, and then it never dies back, despite there being nothing in the queue and few messages coming in overnight.

 

One possible explanation for this would be if the 10 bulk processes all spot a new message in the queue (resetting their counters) but only one processes it, or if the new messages are assigned to bulk threads randomly, meaning that with a 5min lazytime, you only need 1 message in 30s to be able to reset all counters, and one message every second makes it pretty much guaranteed.

 

Does anyone know any details of this, and have any suggested values for these settings in a higher-throughput environment?  It may be worthwhile having a ‘bulk_max_messages’ option, allowing a bulk thread to die after processing a set maximum of messages (protecting us against memory leaks and excessive threads)

 

Steve

 

Steve Shipway

University of Auckland

UNIX Systems Design Team Lead

address@concealed

+64 (9) 3737 599 ext 86487

(GNU Terry Pratchett)

 

Attachment: smime.p7s
Description: S/MIME cryptographic signature