You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Reported in the testing group: it seems like there's a potential performance issue with put_unique. Many writes from Zeek lead to an increasing memory usage over time until the system keels over.
The text was updated successfully, but these errors were encountered:
To clarify, the actual report was about memory bloat. So far, I couldn't reproduce the bloat itself. What I did find is that Broker seems to do fine for throughput, but latency is quite high. This makes roundtrips for put_unique quite slow and Broker can only do ~4-5k put_unique per second with one writer/clone. Adding more writers/clones slows down further. I didn't try "interleaving" writes, though. The setup in the testing groups seems to be calling put_unique non-stop (without waiting for results first). That could explain the bloat if the system simply does more writes than Broker can handle.
Reported in the testing group: it seems like there's a potential performance issue with
put_unique
. Many writes from Zeek lead to an increasing memory usage over time until the system keels over.The text was updated successfully, but these errors were encountered: