Hi Lisa, (adjusting subject, perhaps it draws more attention)
On Mon, 2021-02-15 at 18:26 -0500, Lisa Davidson wrote: > We run multiple application pods on OpenShift(AWS) that talk to an > external > mongo db cluster which spreads out in datacenters(not AWS). > > During a deployment, it normally takes about 10 mins for "Apache > Sling > Repository Startup Thread" to do its thing. What is it doing? Any > idea what > it takes so long? Why is every application pod doing repository > startup? > > Do you recommend attaching a persistent volume to the application > pods? Is > there a way for Sling to tell other pods there is no need to do > repository > startup if it was done already? > First off, do these pods connect to the same MongoDB database, or are they independent? 10 minutes is a long time, even for a remote MongoDB instance. The startup thread does all the initial work of creating the repository, which means starting up Oak (including initial content, indexes, etc ) and running any repository initialisers, such as repoinit statements. Do you have any specific log statements that draw attention? Can you profile the execution and see where the time is spent? Thanks, Robert
