package gridengine-master 8.1.9+dfsg-7build1 failed to install/upgrade: installed gridengine-master package post-installation script subprocess returned error exit status 139

[AlexisDelosSantos]

This happened at 99% installing the debian sience distributed computing packages with:

sudo tasksel install science-distributedcomputing

ProblemType: Package
DistroRelease: Ubuntu 18.04
Package: gridengine-master 8.1.9+dfsg-7build1
ProcVersionSignature: Ubuntu 4.15.0-22.24-generic 4.15.17
Uname: Linux 4.15.0-22-generic x86_64
ApportVersion: 2.20.9-0ubuntu7.1
Architecture: amd64
Date: Wed May 30 20:50:23 2018
ErrorMessage: installed gridengine-master package post-installation script subprocess returned error exit status 139
InstallationDate: Installed on 2018-04-27 (33 days ago)
InstallationMedia: Ubuntu-MATE 18.04 LTS "Bionic Beaver" - Release amd64 (20180426)
Python3Details: /usr/bin/python3.6, Python 3.6.5, python3-minimal, 3.6.5-3
PythonDetails: /usr/bin/python2.7, Python 2.7.15rc1, python-minimal, 2.7.15~rc1-1
RelatedPackageVersions:
 dpkg 1.19.0.5ubuntu2
 apt 1.6.1
SourcePackage: gridengine
Title: package gridengine-master 8.1.9+dfsg-7build1 failed to install/upgrade: installed gridengine-master package post-installation script subprocess returned error exit status 139
UpgradeStatus: No upgrade log present (probably fresh install)

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[Lennart Karssen (l.c.karssen)]

It looks like this bug has been present since Ubuntu 16.10: see bug #1633251

[Jody Simpson (jksimpson)]

I was able to cobble together a fix for this, hopefully someone could use this information to create a more refined solution.

The underlying issue is that /usr/lib/gridengine/spooldefaults.bin will segfault when creating the initial databases. gridengine-client_8.1.9+dfsg-9 from debian sid has the appropriate fixes for this. I copied over spooldefaults.bin, libspoolb.so and libspoolc.so from /usr/lib/gridengine in that package.

After this, gridengine-master will try to start and report:
error opening shared lib "/usr/lib/gridengine/libspoolb.so": /usr/lib/x86_64-linux-gnu/libjemalloc.so.2: cannot allocate memory in static TLS block

This can be resolved by compiling jemalloc with --disable-initial-exec-tls

[Christoph Probst (christophprobst)]

Can confirm that using spooldefaults.bin, libspoolb.so and libspoolc.so from http://ftp.debian.org/debian/pool/main/g/gridengine/gridengine-client_8.1.9+dfsg-9_amd64.deb does workaround the segfault issue in gridengine-master 8.1.9+dfsg-7build1 on Ubuntu 18.04 Bionic.

Make sure to reset /var/spool/gridengine/ if there are slot-related error messages.

Regarding libjemalloc it was sufficiant to symlink version 2 to version 1.
cd /usr/lib/x86_64-linux-gnu
ln -s libjemalloc.so.1 libjemalloc.so.2

I also used the Debian source for 8.1.9+dfsg-9 to build the package locally on Bionic but the result still segfaulted. Possibly lib-related - haven't tried to provide libm and libjemalloc2.

debian dfsg-9 ldd diff:
        libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6
        libjemalloc.so.2 => /usr/lib/x86_64-linux-gnu/libjemalloc.so.2

local build dfsg-9 ldd diff:
        libjemalloc.so.1 => /usr/lib/x86_64-linux-gnu/libjemalloc.so.1

[Callum Lewis-Smith (picoutput.cls)]

I have been grappling with this issue myself over the last couple of weeks and can confirm using the spooldefaults.bin, libspooldb.so and libspoolc.so from the Debian gridengine-client APT package, linked by christophprobst resolved the issue.

However, I have also been able to compile and install the gridengine client from the Debian source tarball for 8.1.9+dfsg-9 and haven't encountered any segfaults.