[Pulp-list] Trying HA/loadbalance for pulp

[joshi dhaval]

Hello,

I am trying to use PULP as loadbalancer ( not exactly HA ) ..

i have 2 regions UK and TK ...

i have installed my PULP server in UK with mongodb database + apache, PULP broker, worker and celery beat ....

Now i have installed second server in TK with PULP broker, apache, worker ( not started celerybeat and resource_manager ) and my DB is pointing to UK for all DB entries ..

i run rsync for /var/lib/pulp ( it's not shared over NFS ) to sync contents in some time intervals ...

i am facing issues while doing uploads ( the error messages are related to celery workers ) so when i try to upload package in repository available in TK, if it has worker processes running it doesnt upload package , same thing happens for Publish ... if i stop worker .. it just works ...... but very random and sometime face issues with PUBLISH ...

can you suggest what processes should running on TK and UK ??  also what FS needs to be shared ??

Now why i am doing this is because upload from UK to TK pulp server for 1GB RPM it takes almost 30Mins to upload package in repository ( while rycn just takes 1 min to transfer file ) ... any suggestion will help to improve it .....




Regards,
DJ