around 15:37PDT I noticed my CDS overview MEDM was not connecting to the models running on h1seih23. The control room TV MEDM screen was not showing this problem, but new MEDM connections were not being established. I was also unable to ping or ssh onto h1seih23.

h1seih23's console is reporting many errors (several per second) of the type:

nf_conntrack: table full, dropping packet.

and sometimes

net_ratelimit: 66 callbacks suppressed

a quick goggle search shows that the internal IP connection table has reached its limit.

Over time new MEDM connections are actually being established. Guardian is not reporting any issues with its connections.

So we are thinking that new channel access connections are flakey, but established ones (like guardian) will continue to work. The IFO team would like to delay a restart of h1seih23 until tomorrow unless it becomes a major issue to IFO locking.