cancel
Showing results for 
Search instead for 
Did you mean: 

how to reduce relpace time

邹英琼_1
Occasional Contributor

how to reduce relpace time

i have a single application in TRU64 cluster 5.1b . when one server is power off immediate ,and another server will replace this applicate .but it cost for a long time .exampe: 121 seconds CNX disk rebuild ,and 60 seconds application
replace .
if a application replace will take for a long time .the total time is :250 seconds .
but my custom would not accept this long time.
i hope to short :121 seconds ,how can i do.???
2 REPLIES
Gary Hansford
Frequent Advisor

Re: how to reduce relpace time

Go to the Tru64 best practices documents at:-
http://h30097.www3.hp.com/docs/best_practices/clus_bps.html

You'll find the document "Tuning the Cluster Transition Time" is what you are looking for. This shows how you can reduce the time (via tweaking of cluster_rebuild_delay).

At this site I have the cluster_rebuild_delay set at 60 for what it is worth.

Cheers

Gary
Johan Brusche
Honored Contributor

Re: how to reduce relpace time


The best practice mentionned in previous entry says the minimum rebuild_delay = 50.

For more recent versions like V5.1B + pk#4
the minimum rebuild_delay = 25.

Some customers, that wanted even less have received a customer specific patch for which the minimum rebuild_delay =15.

A word of caution is in order here.....
Havy ics0 traffic can cause a false rebuild trigger due to the contention between filesystem traffic and cnx manager handshakes. This was observed during backup, where the backup data travelled over the memory channel.

Complementary to the cluster_rebuild_delay you can also increase the frequence of the "check" procedure and decrease the number of startup failures for that particular CAA-service to make it relocate more promptly even when the member on which it runs is OK.

Kind regards,

___ Johan.

_JB_