sahara-all leaks memory

Bug #1532826 reported by Francois Deppierraz
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Sahara
Triaged
Critical
Unassigned

Bug Description

OpenStack kilo deployment using the Ubuntu Cloud Archive under Ubuntu 14.04 and running Sahara with the direct infrastructure engine (as the heat one didn't worked in this setup).

This deployment was done with the official puppet-sahara modules and there are currently only 2 clusters running (4 and 16 nodes). Many test clusters have been launched and deleted before though.

The 'sahara-all' process has been running for 21 days and now consumes around 5.5 GB of RAM.

$ lsb_release -a
No LSB modules are available.
Distributor ID: Ubuntu
Description: Ubuntu 14.04.3 LTS
Release: 14.04
Codename: trusty
$ apt-cache policy sahara-common
sahara-common:
  Installed: 2015.1.0-0ubuntu1~cloud0
  Candidate: 2015.1.0-0ubuntu1~cloud0
  Version table:
 *** 2015.1.0-0ubuntu1~cloud0 0
       1000 http://ubuntu-cloud.archive.canonical.com/ubuntu/ trusty-updates/kilo/main amd64 Packages
        100 /var/lib/dpkg/status
$ top -b -n 1 | grep sahara-all
 1057 sahara 20 0 5876080 5.506g 5376 S 0.0 17.5 552:40.64 sahara-all
$

While trying to understand what all this memory is being used for, I've discovered many instance of SSH private key in the process memory. It seems that ~3 GB of memory (1.7 mio instance at 1.7 KB each) is being used for these keys, this may be a hint on where the memory leaks actually comes from.

$ sudo gcore 1057
[...]
Saved corefile core.1057
$ strings core.1057 | sort | uniq -c | sort -rn | head
1771726 -----END RSA PRIVATE KEY-----
[...]
$ echo $[1771490*1679]
2974331710
$

Revision history for this message
Francois Deppierraz (francois-ctrlaltdel) wrote :
description: updated
Revision history for this message
Evgeny Sikachev (esikachev) wrote :

please add method for deploy openstack to description

Changed in sahara:
status: New → Incomplete
description: updated
Changed in sahara:
status: Incomplete → New
Changed in sahara:
status: New → Confirmed
Changed in sahara:
status: Confirmed → Triaged
importance: Undecided → Critical
Revision history for this message
Sergey Reshetnyak (sreshetniak) wrote :

same problem with sahara-engine

Revision history for this message
Evgeny Sikachev (esikachev) wrote :

bug reproduced on stable/liberty

no longer affects: sahara/mitaka
Changed in sahara:
milestone: mitaka-rc1 → newton-1
no longer affects: sahara/newton
no longer affects: sahara/mitaka
no longer affects: sahara/newton
Changed in sahara:
milestone: none → newton-1
no longer affects: sahara/liberty
Changed in sahara:
assignee: Sergey Reshetnyak (sreshetniak) → nobody
Revision history for this message
Francois Deppierraz (francois-ctrlaltdel) wrote :

Hi Vitaly, Do you have any pointer to commits which solved this issue?

Thanks!

Revision history for this message
Vitalii Gridnev (vgridnev) wrote :

Francois, we didn't came up with a solution of this problem, actually. So, I did reset of assignee for this bug and removed liberty series for awhile (until valid solution will be found)

Changed in sahara:
milestone: newton-1 → none
no longer affects: sahara/mitaka
Changed in sahara:
milestone: none → next
Changed in sahara:
milestone: next → none
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Bug attachments

Remote bug watches

Bug watches keep track of this bug in other bug trackers.