mem=2048 for landscape-server may lead to OOM killer with some periodic tasks

[Nobuto Murata]

It's reproducible with local provider too. It's better to bump the memory constraint to 4096.

How to reproduce:
$ juju deploy landscape-scalable

$ juju run --unit landscape-server/0 -- \
    sudo -u landscape -H /opt/canonical/landscape/scripts/update_security_db.sh

    /opt/canonical/landscape/scripts/update_security_db.sh: line 52: 5574 Done bzcat ${OUTPUT_FILE}
          5575 Killed | ./process-usns /dev/stdin 2>&1
          5576 | pipe_to_syslog $LOGGER_NAME
    Error running /opt/canonical/landscape/scripts/update_security_db.sh: 0
    Check out the syslog output for script update_security_db.sh.

    Aug 14 08:35:15 darkbox kernel: [12340.389945] Memory cgroup out of memory: Kill process 27221 (python) score 174 or sacrifice child
    Aug 14 08:35:15 darkbox kernel: [12340.389960] Killed process 27221 (python) total-vm:719436kB, anon-rss:348344kB, file-rss:0kB, shmem-rss:0kB
    Aug 14 08:35:15 darkbox kernel: [12340.432737] oom_reaper: reaped process 27221 (python), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB