Re: PG10 - Out of shared memory / invalid memory alloc request size

I remember an old issue I had with postgres and shared memory being set too low by the distro I was using.

What's the value of "/proc/sys/kernel/shmmax"? Try setting to 18446744073709551615

Regards,

Alvaro Aguayo
Operations Manager
Open Comb Systems E.I.R.L.

Office: (+51-1) 3377813 | Mobile: (+51) 995540103 | (+51) 954183248
Web: www.ocs.pe

----- Original Message -----
From: "Arjun Ranade" <ranade(at)nodalexchange(dot)com>
To: "pgsql-admin" <pgsql-admin(at)postgresql(dot)org>
Sent: Tuesday, 14 November, 2017 14:44:08
Subject: [ADMIN] PG10 - Out of shared memory / invalid memory alloc request size

We recently upgraded a database that a number of developers use from
Postgres 9.4 to 10. Our memory parameters are unchanged.

We are seeing a lot of errors that look like this in the postgres log:

< 2017-11-14 13:13:59 EST [144172] : [1-1]
user=[unknown],db=qa16,remote=192.168.104.22(35238) > FATAL: no PostgreSQL
user name specified in startup packet
< 2017-11-14 13:52:22 EST [6273] : [1-1]
user=deployment_app,db=qa5,remote=192.168.104.83(45384) > ERROR: out of
shared memory
< 2017-11-14 14:17:53 EST [6200] : [3-1]
user=service_tracker_app,db=qa16,remote=192.168.104.22(46996) > ERROR:
invalid memory alloc request size 34359738368

Anecdotally it seems that these errors are happening when
powa_take_snapshot() (Postgres workload analyzer) is running as well but I
can't confirm that.

Does anyone know how to debug these types of errors? And once the db is in
this state, how do we resolve sessions reporting "Error: invalid memory
alloc request size" - killing the sessions and having them reconnect does
not fix the issue.

Thanks,
Arjun