DBeaver Experiencing timeouts while connecting to New Linux PostgreSql server

On 5/27/26 22:27, Srinivas Kumar wrote:

Hello,

My company has a new Linux production server of Postgresql 18. While
connecting to it in DBeaver, users are experiencing timeouts with broken
pipe
errors in logs. It connects randomly, 1 in every 3-4 attempts. Since this is
new servers in high available fashion with another standby server, there is
very little traffic and max connections are set to 300. How do we find
RCA for
this? Let me know if any additional details are required. I have checked all
security information and telnet, and it works just fine in CMD or a
terminal.

Looking forward to hearing from you.

There not much information, so it's hard to give you good advice. We
don't even know how the users connect - is it directly or through a
pool/proxy of some sorts, or what? What logs have messages about broken
pipes - is that the Postgres server log or some other log?

You'll have to do the usual thing people do during RCA analysis, i.e. go
step by step over all the components on the connection (between the
client and the server), figure out where exactly it fails and why.

My guess is something is dropping the connection for whatever reason
(might be a proxy, might be a VPN, ... who knows?).

regards

--
Tomas Vondra

Hello Tomas,

Thanks for reaching out! For more details, users connect directly to
PostgreSQL using the DBeaver tool. Please review the attached error
messages in the PostgreSQL server logs, and the timeout error on the
EDWDBPRDL01A server. We have another dev server in Windows, which works
absolutely fine without this issue, as you can review. Please let me know
if any more specific details are required.

Thanks,
Sri.

On Wed, May 27, 2026 at 5:37 PM Tomas Vondra <tomas@vondra.me> wrote:

On Wed, May 27, 2026 at 4:19 PM Srinivas Kumar <srinivasnerusu@gmail.com>
wrote:

Hello Tomas,

Thanks for reaching out! For more details, users connect directly to
PostgreSQL using the DBeaver tool. Please review the attached error
messages in the PostgreSQL server logs, and the timeout error on the
EDWDBPRDL01A server. We have another dev server in Windows, which works
absolutely fine without this issue, as you can review. Please let me know
if any more specific details are required.

From the logs you sent I don't see anything that jumps out.

- 58 lines total
- 6 connection authenticated
- 6 connection authorized

Only one problem: FATAL 08006 connection to client lost on PID 1994906
(pgAdmin -> EDW, 18:56:11). Preceded by LOG 08006 - could not send data to
client: Broken pipe. pgAdmin dropped the TCP socket while the session was
idle. Postgres detected it when it tried to write back. Client-side
disconnect, not a server error.

Everything else is normal: SETs, idle state transitions, and clean
disconnections. No auth failures, no query errors, no locks, no timeouts
recorded server-side.

Roberto

In our company all team members face this broken pipe , client side
disconnect while using DBeaver. Any fixes for it? As we have same tool
working fine for dev server.

On Wed, 27 May 2026 at 7:49 PM, Roberto Mello <roberto.mello@gmail.com>
wrote:

Any further tips/ troubleshooting needed for this issue? Please let me know.

Thanks,
Sri.

On Wed, May 27, 2026 at 8:14 PM Srinivas Kumar <srinivasnerusu@gmail.com>
wrote:

Sri,

I already gave you a couple general suggestions how you might
investigate this. What have you tried since then? Have you checked the
components on the path between the server and the clients? Have you foun
any new clues? Have you tried connecting using a different client
(psql/pgadmin/...)?

If not, I doubt anyone can give you any sensible advice - not
considering the lack of information. All we have is your "it
disconnects" description, and 30 lines from a server log that shows
nothing about clients disconnecting.

regards

On 6/1/26 16:57, Srinivas Kumar wrote:

Any further tips/ troubleshooting needed for this issue? Please let me know.

Thanks,
Sri.

On Wed, May 27, 2026 at 8:14 PM Srinivas Kumar <srinivasnerusu@gmail.com
<mailto:srinivasnerusu@gmail.com>> wrote:

In our company all team members face this broken pipe , client side
disconnect while using DBeaver. Any fixes for it? As we have same
tool working fine for dev server. 

On Wed, 27 May 2026 at 7:49 PM, Roberto Mello
<roberto.mello@gmail.com <mailto:roberto.mello@gmail.com>> wrote:

On Wed, May 27, 2026 at 4:19 PM Srinivas Kumar
<srinivasnerusu@gmail.com <mailto:srinivasnerusu@gmail.com>> wrote:

Hello Tomas,

Thanks for reaching out! For more details, users connect
directly to PostgreSQL using the DBeaver tool. Please review
the attached error messages in the PostgreSQL server logs,
and the timeout error on the EDWDBPRDL01A server. We have
another dev server in Windows, which works absolutely fine
without this issue, as you can review. Please let me know if
any more specific details are required. 

From the logs you sent I don't see anything that jumps out.

  - 58 lines total
  - 6 connection authenticated
  - 6 connection authorized

Only one problem: FATAL 08006 connection to client lost on PID
1994906 (pgAdmin -> EDW, 18:56:11). Preceded by LOG 08006 -
could not send data to client: Broken pipe. pgAdmin dropped the
TCP socket while the session was idle. Postgres detected it when
it tried to write back. Client-side disconnect, not a server error.

Everything else is normal: SETs, idle state transitions, and
clean disconnections. No auth failures, no query errors, no
locks, no timeouts recorded server-side. 

Roberto

--
Tomas Vondra

Hello Tomas,

Yes, I have tried to run several commands from linux as below

Connectivity Checks
1. Check server reachability
Bash
ping <postgres_host>
2. Check PostgreSQL port connectivity
Bash
nc -zv <postgres_host> 5432
OR
Bash
telnet <postgres_host> 5432
3. Check connection timing
Bash
time nc -zv <postgres_host> 5432
4. Test PostgreSQL login timing
Bash
time psql -h <postgres_host> -U <user> -d <database>
PostgreSQL Checks
5. Check active sessions
SQL
SELECT count(*) FROM pg_stat_activity;
6. Check max connections
SQL
SHOW max_connections;
7. Check idle/stuck sessions
SQL
SELECT pid, usename, state, backend_start, state_change
FROM pg_stat_activity
ORDER BY state_change;
Linux/Socket Checks
8. Check TCP sessions
Bash
ss -ant | grep 5432

9. Monitor logs during issue time
Bash
tail -f /var/log/postgresql/postgresql*.log

10. Run continuous connectivity test
Bash
while true
do
date
nc -zv <postgres_host> 5432
sleep 3
done

All worked fine except for scenarios connecting via DBeaver.
Also, psql and pgadmin connections are fast, except that pgadmin, it asks
to re-enter the user/login password after a few minutes.
My company has a Windows dev server, which doesn't experience any such
issues from DBeaver/pgadmin, etc.

Let me know if any specific troubleshooting steps are recommended to test
further.

Thanks,
Sri.

On Mon, Jun 1, 2026 at 7:07 PM Tomas Vondra <tomas@vondra.me> wrote: