Hi Jan Thank you for the reply.
Version 3.1. is definite, on both Windows and Linux. Can you point me toward notes on remsh? Very happy to try to reproduce it, though we haven't found reliable steps to do that yet. Thanks Andrew On Tue., 25 Apr. 2023, 20:59 Jan Lehnardt, <[email protected]> wrote: > Heya, > > The last time we fixed something along these lines was in 2019: > > https://github.com/apache/couchdb/pull/2195/files > > Which predates 3.1.0: > > https://blog.couchdb.org/2020/05/06/3-0-1-3-1-0/ > > Are you 100% sure about your version? > > A quick look does not suggest we have introduced this pattern again > since. > > But CouchDB should behave fine in your scenario. > > Can you try and remsh into the node and see which erlang processes > are hogging the CPU? > > Also we should move this to the bug tracker if you can reproduce this > on 3.3.2. > > Best > Jan > — > > On 25. Apr 2023, at 05:53, Andrew E <[email protected]> wrote: > > > > Hi all, > > > > Firstly, sorry for posting this previously on the dev channel. > > > > Our issue... > > > > We use CouchDB (3.1.0) on Windows and Linux. > > > > We have noticed that on Windows startup and wake from hibernate, Couch > > sometimes "goes bananas" for a period of time. > > > > By that I mean it shows a CPU load of >30% for 10-20 minutes on a > > reasonably spec'd Windows developer laptop. After that period, erl.exe > goes > > to a normal "almost invisible" CPU level. > > > > The active task list is shown as empty in Fauxton. > > > > In Production the Windows instances are on vehicles that are not > reachable > > on the network for hours at a time while collecting data. They replicate > to > > the Linux servers when the network is present. Those computers are > > underpowered compared to development laptops, which means they are barely > > usable when Couch grinds like this. > > > > Any ideas what could be going on or how to find out? > > > > Thank you! > > Andrew > >
