Scalability issues

[shroffk]

FRIB (@berryma4) reported that in situations where many IOC's each with 10k's of pvs are started at the same time the receiver crashes.
It is due to memory overflows with the receiver bufferring the record lists from all the IOC's.

[mdavidsaver]

The main tool I made for bounding peak memory usage is maxActive, which restricts the number of IOCs allowed to concurrently dump record lists.

recsync/server/demo.conf

Lines 42 to 44 in 7ebed83

	# Maximum concurrent "active" clients
	# to allow.
	#maxActive = 20

@berryma4 Have you tried lower this from the default? Also, did you save any stack traces from the crashes you saw?

[mdavidsaver]

@berryma4 ping. (maybe @Dylan171 knows the story)

[anderslindho]

Turns out maxActive never actually was used, or that it had been broken at some point in history - see 956ced8. I wonder if we now can/should close this ticket?