BUG #13641: PostgreSQL (pg_ctl.exe) conflicts with GPClient.exe - Mailing list pgsql-bugs

From it@ramsoft.com
Subject BUG #13641: PostgreSQL (pg_ctl.exe) conflicts with GPClient.exe
Date
Msg-id 20150925234100.3022.96933@wrigleys.postgresql.org
Whole thread Raw
List pgsql-bugs
The following bug has been logged on the website:

Bug reference:      13641
Logged by:          Nick Childs
Email address:      it@ramsoft.com
PostgreSQL version: 9.1.18
Operating system:   Windows Server 2012 R2
Description:

I opened this bug over a month ago, and it has not been touched - creating a
duplicate in the hopes that someone will do something about this:
http://www.postgresql.org/message-id/20150829014242.2090.78049@wrigleys.postgresql.org

We use PostgreSQL version 9.1.18 as part of a Pentaho deployment in our
software. We have a service called, "RamSoft Essence Solution Repository
Service," which runs the following command:

C:/Program Files/RamSoft/Essence/postgresql/bin/pg_ctl.exe runservice -N
"RamSoft Essence Solution Repository" -D "C:/Program
Files/RamSoft/Essence/postgresql/data"

About 6 months ago, we believe we were hit with the following bug in a
previous version of PostgreSQL, which caused major login/reboot delays due
to a hung gpclient process when Registry/Printer/Shortcut GPO's are applied
to the server running PostgreSQL:
http://www.postgresql.org/message-id/CAEB4t-NpXGiP5Bqvv3P+d+x=V4BqE+Awg+G7ennBn8icPXep_g@mail.gmail.com
- it would delay the ability to log in to the server or reboot for over 40
minutes, unless gpclient.exe was force killed or the service was disabled
before reboot.

Our Development team provided a procedure to upgrade PostgreSQL to 9.1.18,
which does appear to have successfully resolved the login/reboot delay
issues with our domain-joined machines running PostgreSQL for all GPO's
except a shortcut GPO that is applied in one of our datacenters. For some
reason, this specific shortcut policy still triggers the bug after 9.1.18
upgrade.

However, we have noticed an additional behavior/problem between pg_ctl.exe
and gpclient.exe in domain environments with Registry/Printer/Shortcut
GPO's
applied:

- Reboot VM; no extended delays are experienced
- Log in to VM: no extended delays are experienced
- Open Administrative Command Prompt, and run: gpupdate /force - command
hangs indefinitely; never completes
- Stop Essence Solution Repository Service (pg_ctl.exe)
- Re-run gpupdate /force - completes very quickly, and successfully
- Start Essence Solution Repository Service
- Re-run gpupdate /force - completes very quickly, and successfully
- Perform reboot
- gpupdate /force hangs again, until we stop/start Essence Solution
Repository Service

Disabling Essence Solution Repository Service before the reboot allows
gpupdate /force to complete quickly and successfully immediately after
logging in. Starting the service after login allows gpupdate /force to
complete. Setting the service to delayed start does NOT allow gpupdate
/force to complete unless it is stopped/started before running the command.

So, it seems that there is still some sort of conflict between the initial
state of Essence Solution Repository Service (pg_ctl.exe) and gpclient.exe
-
it needs to be stopped before gpupdate /force can run successfully, for
some
reason.

- Is this a known issue on 9.1.18?
- Is this reproducible for you?
- We also need to know why version 9.1.18 fixed the delays for all GPO
settings except for Shortcut-related policies.
- Is there anything we can do to help you solve this problem?

Please let us know if you have any questions.

pgsql-bugs by date:

Previous
From: Tom Lane
Date:
Subject: Re: BUG #13640: Delete isn't using available Index Only Scan
Next
From: amir.rohan@mail.com
Date:
Subject: BUG #13642: no backup_label file in PG_DATA after pg_stop_backup();