Unable to use qcow2 disk image larger than system memory
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
ironic-python-agent |
Won't Fix
|
Wishlist
|
Jonathan Provost |
Bug Description
Background:
* That was when using the agent driver
* Node with 4GB RAM
* 10GB qcow2 image
* CoreOS ramdisk
Steps to reproduce:
* Setup node in Ironic
* Deploy node with large qcow2 image
Actual behavior:
The machine becomes unresponsive and crashes. Deployment fails without too much indication as to what part failed.
Expected behavior:
Large qcow2 images are correctly imaged into the machine
*OR* I quickly get an error message that this image cannot be installed on the machine because it is too big.
Analysis:
Any image to be deployed (besides RAW images when using streaming=True) are downloaded to a temporary location, /tmp. The image is then checksummed and passed to write_image.sh which then calls qemu-img for writing to disk.
Because /tmp is backed by the root tmpfs, when the image is larger than the amount of available space, the image fills up the memory and the server crashes.
information type: | Public → Public Security |
information type: | Public Security → Public |
affects: | ironic → ironic-python-agent |
Changed in ironic-python-agent: | |
assignee: | Félix Bouliane (fbouliane) → Jonathan Provost (jprovost-sh) |
Changed in ironic-python-agent: | |
importance: | Undecided → Wishlist |
tags: | added: rfe |
Fix proposed to branch: master /review. openstack. org/428410
Review: https:/