Re: BUG #18315: Segmentation fault on connection (repmgr, psql)

Another update and fun fact.
After a few days of database servers being idle, I no longer see these
errors.

For the sake of additional testing, I rebooted one of the machines in the
cluster and immediately started seeing errors again.

Looks more like an OS issue than PostgreSQL itself, but I have no idea
where to look. The weird thing is that the issue always happens on
libkrb5.so.3.3.
I'm completely lost here.

I can confirm that the underlying hardware (on the hypervisor) is OK.

Any additional ideas would be of help. Thanks

Ivan Kalafatić

On Tue, 30 Jan 2024 at 21:22, Ivan Kalafatić <ikalafat(at)gmail(dot)com> wrote:

> Additional info: in the meantime, I've tried adding require_auth=none
> and gssencmode=disable connection string parameters (separately, not
> together at once) for the repmgr/client connections that are "trusted"
> through pg_hba.conf, but the problem persisted.
>
> On Tue, 30 Jan 2024 at 16:17, PG Bug reporting form <
> noreply(at)postgresql(dot)org> wrote:
>
>> The following bug has been logged on the website:
>>
>> Bug reference: 18315
>> Logged by: Ivan Kalafatić
>> Email address: ikalafat(at)gmail(dot)com
>> PostgreSQL version: 16.1
>> Operating system: Debian 12
>> Description:
>>
>> Hi,
>>
>> I am intermittently getting following errors (taken from var/log/syslog),
>> on
>> pg_basebackup, repmgrd and psql (so far)
>>
>>
>> repmgr[1147164] general protection fault ip:7fb9950981b5 sp:7ffecf4418f8
>> error:0 in libkrb5.so.3.3[7fb99506a000+5f000]
>> also this:
>>
>> pg_basebackup[4191295] general protection fault ip:7f344e9771b5
>> sp:7ffe8c625c28 error:0 in libkrb5.so.3.3[7f344e949000+5f000]
>>
>> and this
>>
>> psql[4045720] general protection fault ip:7f89130931b5 sp:7fffd40a0268
>> error:0 in libkrb5.so.3.3[7f8913065000+5f000]
>>
>> Looks like that the error is related to GSSAPI/kerberos auth? I have tried
>> replacing "host" type within pg_hba.conf to "hostnogssenc", but problem
>> persists.
>>
>> Running Debian 12 with all updates, repmgr, timescaledb extensions (fresh
>> install). PostgreSQL 16.1 (Debian 16.1-1.pgdg120+1)
>> Looks like the server itself is not affected (isn't crashing etc).
>>
>> Is there anything that I could do to disable GSSAPI/kerberos and/or
>> circumvent this issue? Also, how can I provide more info for this issue to
>> be resolved.
>>
>> Thanks.
>>
>>