Re: New PG14 server won't start with >2GB shared_buffers

Tom,

Got a chance to work on this today. Here is what I’m getting. I used the same command that is used when the config has small memory.

$ /usr/lib/postgresql/14/bin/postgres -D /var/lib/postgresql/14/<dir> --config-file=/etc/postgresql/14/appdb/postgresql.conf --listen_addresses=0.0.0.0 --port=5432 —cluster_name=<name> --wal_level=logical --hot_standby=on --max_connections=150 --max_wal_senders=10 --max_prepared_transactions=0 --max_locks_per_transaction=64 --track_commit_timestamp=off --max_replication_slots=10 --max_worker_processes=12 --wal_log_hints=on
FATAL: could not map anonymous shared memory: Cannot allocate memory
HINT: This error usually means that PostgreSQL's request for a shared memory segment exceeded available memory, swap space, or huge pages. To reduce the request size (currently 21967716352 bytes), reduce PostgreSQL's shared memory usage, perhaps by reducing shared_buffers or max_connections.
LOG: database system is shut down

$ /usr/lib/postgresql/14/bin/postgres -D /var/lib/postgresql/14/<dir> --config-file=/etc/postgresql/14/appdb/postgresql.conf --listen_addresses=0.0.0.0 --port=5432 —cluster_name=<name> --wal_level=logical --hot_standby=on --max_connections=150 --max_wal_senders=10 --max_prepared_transactions=0 --max_locks_per_transaction=64 --track_commit_timestamp=off --max_replication_slots=10 --max_worker_processes=12 --wal_log_hints=on
FATAL: could not create shared memory segment: Cannot allocate memory
DETAIL: Failed system call was shmget(key=2359323, size=21967716352, 03600).
HINT: This error usually means that PostgreSQL's request for a shared memory segment exceeded your kernel's SHMALL parameter. You might need to reconfigure the kernel with larger SHMALL.
The PostgreSQL documentation contains more information about shared memory configuration.
LOG: database system is shut down

$ ulimit -v
unlimited

I’m not sure about the cgroup stuff. I installed lscgroup and this is what is shows. (BTW, we are using Patroni to control PostgreSQL, but it does not appear to be part of the problem since I’ve duplicated the errors without it):
$ lscgroup
devices:/
devices:/user.slice
devices:/system.slice
devices:/system.slice/irqbalance.service
devices:/system.slice/system-systemd\x2dfsck.slice
devices:/system.slice/syslog-ng.service
devices:/system.slice/systemd-networkd.service
devices:/system.slice/systemd-udevd.service
devices:/system.slice/cron.service
devices:/system.slice/oddjobd.service
devices:/system.slice/sys-fs-fuse-connections.mount
devices:/system.slice/sys-kernel-config.mount
devices:/system.slice/networkd-dispatcher.service
devices:/system.slice/sys-kernel-debug.mount
devices:/system.slice/certmonger.service
devices:/system.slice/accounts-daemon.service
devices:/system.slice/swapfile.swap
devices:/system.slice/numad.service
devices:/system.slice/systemd-journald.service
devices:/system.slice/unattended-upgrades.service
devices:/system.slice/sensu-client.service
devices:/system.slice/ssh.service
devices:/system.slice/dev-mqueue.mount
devices:/system.slice/rpc-gssd.service
devices:/system.slice/vnstat.service
devices:/system.slice/var-lib-postgresql.mount
devices:/system.slice/rpcbind.service
devices:/system.slice/chrony.service
devices:/system.slice/sssd.service
devices:/system.slice/proc-sys-fs-binfmt_misc.mount
devices:/system.slice/run-rpc_pipefs.mount
devices:/system.slice/autofs.service
devices:/system.slice/patroni.service
devices:/system.slice/consul.service
devices:/system.slice/telegraf.service
devices:/system.slice/dev-hugepages.mount
devices:/system.slice/dbus.service
devices:/system.slice/system-getty.slice
devices:/system.slice/systemd-logind.service
cpuset:/
cpu,cpuacct:/
cpu,cpuacct:/user.slice
cpu,cpuacct:/user.slice/user-5088.slice
cpu,cpuacct:/system.slice
memory:/
pids:/
pids:/user.slice
pids:/user.slice/user-5088.slice
pids:/user.slice/user-5088.slice/user(at)5088(dot)service
pids:/user.slice/user-5088.slice/session-267.scope
pids:/user.slice/user-5088.slice/session-269.scope
pids:/user.slice/user-5088.slice/session-88.scope
pids:/system.slice
pids:/system.slice/irqbalance.service
pids:/system.slice/system-systemd\x2dfsck.slice
pids:/system.slice/syslog-ng.service
pids:/system.slice/systemd-networkd.service
pids:/system.slice/systemd-udevd.service
pids:/system.slice/cron.service
pids:/system.slice/oddjobd.service
pids:/system.slice/sys-fs-fuse-connections.mount
pids:/system.slice/sys-kernel-config.mount
pids:/system.slice/networkd-dispatcher.service
pids:/system.slice/sys-kernel-debug.mount
pids:/system.slice/certmonger.service
pids:/system.slice/accounts-daemon.service
pids:/system.slice/swapfile.swap
pids:/system.slice/numad.service
pids:/system.slice/systemd-journald.service
pids:/system.slice/unattended-upgrades.service
pids:/system.slice/sensu-client.service
pids:/system.slice/ssh.service
pids:/system.slice/dev-mqueue.mount
pids:/system.slice/rpc-gssd.service
pids:/system.slice/vnstat.service
pids:/system.slice/var-lib-postgresql.mount
pids:/system.slice/rpcbind.service
pids:/system.slice/chrony.service
pids:/system.slice/sssd.service
pids:/system.slice/proc-sys-fs-binfmt_misc.mount
pids:/system.slice/run-rpc_pipefs.mount
pids:/system.slice/autofs.service
pids:/system.slice/patroni.service
pids:/system.slice/consul.service
pids:/system.slice/telegraf.service
pids:/system.slice/dev-hugepages.mount
pids:/system.slice/dbus.service
pids:/system.slice/system-getty.slice
pids:/system.slice/system-getty.slice/getty(at)tty1(dot)service
pids:/system.slice/systemd-logind.service
rdma:/
hugetlb:/
perf_event:/
blkio:/
freezer:/
net_cls,net_prio:/

Thanks,

Chris Hoover
Senior DBA
AWeber.com
Cell: (803) 528-2269
Email: chrish(at)aweber(dot)com

> On Feb 25, 2023, at 11:35 AM, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
>
> Evan Rempel <erempel(at)uvic(dot)ca> writes:
>> Bear in mind that if you are using systemd to start postgres, then these
>> user limits may not apply.
>
> Yeah, it seems likely that the PG server is being started under
> more-restrictive limits than what these manual reports suggest.
> It would be useful to try logging in as the Postgres OS user and
> manually starting the server -- just do "postgres &" and see what
> happens. (If it does start, "pg_ctl stop" can be used to shut it
> down again, or you can manually send SIGTERM to the postmaster
> process.)
>
> I tried to reproduce the problem by intentionally setting
> "ulimit -v" too small for my PG settings, and I got error messages
> that were similar but not identical to what Chris reported.
> (I think the sysv case failed at shmat() not shmget().) So it's
> probably not ulimit per se that's responsible. But if the
> server is being started under systemd, then I can entirely
> believe that systemd has some poorly-documented feature that
> sets additional limits for daemon processes.
>
> I'm still wondering about cgroups, too.
>
> regards, tom lane