hellow.
I have a sync + couch for replicating changes by product. and I am having questions with performance and monitoring.
at the moment indicators
sgw_replication_pull_num_pull_repl_active_continuous ~ 200
sgw_database_num_doc_writes 60,000 / 6hours
and situations arise when clients cannot receive changes for about 6 … 24 hours. if you delete the client database and then start replication again, this also takes a long time (at times of peak load).
I decided to monitor the indicator
sgw_replication_pull_num_pull_repl_active_continuous - sgw_replication_pull_num_pull_repl_caught_up
but it is absurdly small, at peak times (no more than 5), although I know for sure that I have more than 100 client bases, with the latest changes not received.
I decided to look at sgw_replication_pull_rev_send_latency, but the values are 300 * 10 ^ 12.
can you tell me how I can:
find out how many customer bases have not received the latest changes at the moment?
how long does it take on average to receive changes? those. what is the delay.
and where should I dig further (why so slowly)?