Re: Understanding PostgreSQL installer debug log

[Ramesh Maddi <mymail.ramesh@xxxxxxxxx>]

Our product is certified along with PostgreSQL 9.6.6 only. We cannot use further version of PostgreSQL at this time in production until it is certified for our product( which is integrated may other components)  and is officially released.

To elaborate actual issue with client in installing PostgreSQL, it is failing to install on only one client's server RHEL 7.6. Below is the error....

performing post-bootstrap initialization ... Failed to initialize the database cluster with initdb with the error "invalid byte sequence for encoding "UTF8"

Script stderr:
 FATAL:  invalid byte sequence for encoding "UTF8": 0xeb 0x2f 0xdb
child process exited with exit code 1
initdb: removing contents of data directory "/home/postgres/9.6/data"

Error running /home/postgres/9.6/installer/server/initcluster.sh "postgres" "postgres" "/home/postgres/9.6" "/home/postgres/9.6/data" 5432 DEFAULT: FATAL:  invalid byte sequence for encoding "UTF8": 0xeb 0x2f 0xdb
child process exited with exit code 1
initdb: removing contents of data directory "/home/postgres/9.6/data"
Problem running post-install step. Installation may not complete correctly
 The database cluster initialisation failed.
Problem running post-install step. Installation may not complete correctly
 The database cluster initialisation failed.

Initially we thought the issue would be with client's VMs locale settings. But client's VMs locale is set to en_US.utf8 format which is expected and default for PostgreSQL.

This issue is only happening for this client only. With this we are suspecting the issue would be client's OS settings but not sure about what settings need to be modified. In order to further debug the issue, I am adding "debug" options to PostgreSQL EDB installer. But the debug log is fully encoded/encrypted in unknown format. We have client session in early next week. Before that we need to identify the debug steps to resolve the issue.

Coming to Luca's question, I understand that using Yum/RPM is better way of installation. That would be our future plan. But for this particular issue, we need to identify debug steps. 

Thanks for your insight.

-- Ramesh

On Fri, Aug 9, 2019 at 4:33 PM Benedict Holland <benedict.m.holland@xxxxxxxxx> wrote:

Does the problem go away if you install pg11? Are the drives you are getting your logs from encrypted? 

Thanks,
~Ben

On Fri, Aug 9, 2019, 3:17 AM Luca Ferrari <fluca1978@xxxxxxxxx> wrote:

On Fri, Aug 9, 2019 at 8:46 AM Ramesh Maddi <mymail.ramesh@xxxxxxxxx> wrote:
> ./postgresql-9.6.6-1-linux-x64.run --installer-language en --serviceaccount postgres --servicename postgresqld  --datadir "/home/postgres/" --prefix  "/home/postgres" --superpassword 1234 --serverport 5432 --debuglevel 4 --debugtrace ./postgresql-debug.log --mode unattended
>

I suspect this is a dump produced by Qt used by EDB installer, maybe
you should ask support to them for this:

</dump>
    <installerVersion>18.8.0</installerVersion>
    <platformInfo>Linux 4.18.0-25-generic x86_64</platformInfo>
</errorDump>

Please note that, my case, I was able to get a dump immediatly because
the data directory did not exist. Why don't you use at least attended
mode to see if it is something as trivial as in my case?

Beside, is there a specific reason why you are not using
distro-specific packages? See the note here
<https://www.enterprisedb.com/downloads/postgres-postgresql-downloads>.
Luca