just fixed an issue with cybre.space's sidekiq worker that was causing it to frequently (30-60s) restart, interrupting longer-running tasks and possibly causing posts to get lost from feeds. hopefully things should be a bit more stable now
Follow
mastodon administration, technical details
if you run mastodon on ubuntu or debian with jemalloc and you notice sidekiq crashing frequently, it may be related to this issue with FFI: https://github.com/mastodon/mastodon/issues/15751
the fix in our case was to add LD_PRELOAD=/usr/lib/x86_64-linux-gnu/libjemalloc.so to the systemd unit files, so that all native (non-ruby) code could use the same memory allocation system
