Increasing the amount of RAM used for /tmp isn't going to be a solution, especially for smaller memory systems -- you'll just run into other problems with the oom killing off parts of the system or the installer.

Do you have an example kickstart and set of repos that demonstrates the problem?

You should also retry this with a current 7.3 nightly. In 7.2 we started using zram (compressed ram) for swap which should help the situation. It is used automatically for systems with up to 2GB of ram and can be turned on for systems with > 2GB RAM by passing inst.zram=on

Proposed patch moving /tmp/yum.cache to /mnt/sysimage/var/tmp/yum.cache after partitioning.

https://github.com/rhinstaller/anaconda/pull/626

This should help with memory usage during the installation, but I don't think there is much that can be done before that. The problem is we need all that metadata for the depsolve and there just isn't any other place to put it.

I've found problem with the commit mentioned in comment 7.

In installed system, there is /mnt/sysimage/var/tmp/yum.cache directory, this is probably caused by following line (note root=iutil.getSysroot() and cache_dir=new_cache already containing /mnt/sysimage/var/tmp/yum.cache):
self._resetYum(root=iutil.getSysroot(), keep_cache=True, releasever=releasever, cache_dir=new_cache)

Moving this to ASSIGNED, since this should be easy to fix and this bug is targeted for RC.

Note:
The directory should be removed by following lines (already in commit) once the cache is in correct place.
# Cleanup temporary yum.cache on disk
if os.path.isdir(iutil.getSysroot()+"/var/tmp/yum.cache"):
    shutil.rmtree(iutil.getSysroot()+"/var/tmp/yum.cache")

Please attach the logs, there should be no way for that block of code to be skipped.

Note that I am unable to reproduce this failure. At the end of installation there is no yum.cache directory in /mnt/sysimage/var/tmp, and it is also not there on reboot. Please make sure you are using the correct version of anaconda. My test was done with anaconda-21.48.22.81-1 from the current nightly.

Sorry, for confusion, I wasn't clear enough.
There is /mnt/sysimage/mnt/sysimage/var/tmp/yum.cache (from anaconda environment), so /mnt/sysimage/var/tmp/yum.cache in installed system.
It's empty directory, but it's still there.
Also, this seems to be 100% reproducible, I've tried interactive, kickstart and installations on different architectures, it was always there.

Attaching logs from installed system.

Created attachment 1189905 [details]
anaconda.log

Created attachment 1189906 [details]
ifcfg.log

Created attachment 1189907 [details]
journal.log

Created attachment 1189908 [details]
ks-script-rRevgC.log

Created attachment 1189909 [details]
ks-script-SQSu2H.log

Created attachment 1189910 [details]
packaging.log

Created attachment 1189911 [details]
program.log

Created attachment 1189912 [details]
storage.log

Created attachment 1189913 [details]
syslog

https://github.com/rhinstaller/anaconda/pull/735

Anaconda moves yum.cache used during installation to disk right after partitioning phase - this help to avoid depletion of /tmp filesystem.

Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.

For information on the advisory, and where to find the updated
files, follow the link below.

If the solution does not work for you, open a new bug report.

https://rhn.redhat.com/errata/RHEA-2016-2158.html