Bug 505263
Summary: | virtual machine installation very slow if "allocate entire virtual disk" not checked | ||
---|---|---|---|
Product: | Red Hat Enterprise Linux 5 | Reporter: | Simon Xu <hxunix> |
Component: | virt-manager | Assignee: | Cole Robinson <crobinso> |
Status: | CLOSED NOTABUG | QA Contact: | Virtualization Bugs <virt-bugs> |
Severity: | medium | Docs Contact: | |
Priority: | low | ||
Version: | 5.3 | CC: | clalance, xen-maint |
Target Milestone: | rc | ||
Target Release: | --- | ||
Hardware: | i686 | ||
OS: | Linux | ||
Whiteboard: | |||
Fixed In Version: | Doc Type: | Bug Fix | |
Doc Text: | Story Points: | --- | |
Clone Of: | Environment: | ||
Last Closed: | 2009-06-11 09:08:09 UTC | Type: | --- |
Regression: | --- | Mount Type: | --- |
Documentation: | --- | CRM: | |
Verified Versions: | Category: | --- | |
oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
Cloudforms Team: | --- | Target Upstream Version: | |
Embargoed: |
Description
Simon Xu
2009-06-11 09:01:58 UTC
In point of fact, this is why we have the "Allocate entire disk now" checked by default. The problem is that when you uncheck this box, a sparse file is created. Any data that needs to be written to the device needs to do a full metadata update, which makes formatting and other operations extremely slow. Switching to another filesystem like ext4 *may* improve the situation, but the basic recommendation is always to use non-sparse files. I'm going to close this as NOTABUG, since it's a known performance problem with sparse files. Chris Lalancette But on x86_64 RHEL5.3, checking "Allocate entire disk now" or not does not make a difference, it is very fast in both ways. (In reply to comment #2) > But on x86_64 RHEL5.3, checking "Allocate entire disk now" or not does not make > a difference, it is very fast in both ways. Is this on the same hardware? I've found that the choice of hardware can affect sparse file performance. However, if it's on exactly the same hardware, then that would be interesting to know. Chris Lalancette Other factors that will impact the performance are reuse of an existing file for a new installation, or use of a block device or LVM volume instead of a file. If the performance is markedly different between 32- and 64-bit in like-for-like testing --- installing to a brand new file on the same filesystem --- then that is definitely something we need to investigate. I have found that, on two x86_64 machines, both 32- and 64-bit RHEL5.3 (2.6.18-128.el5) have the problem, i.e. it'll be very very slow during "write super block" if "allocate entire virtual disk now?" not checked, and during "install package", there is no response. |