Swap space allocation for large memory systems needs improvement
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
partman-auto (Ubuntu) |
Fix Released
|
Medium
|
Unassigned | ||
Precise |
Won't Fix
|
Undecided
|
Unassigned |
Bug Description
The recipe used for automatic swap space allocation during install this is:
100% 512 200% linux-swap
$lvmok{ }
method{ swap }
format{ } .
Where we currently allocate a minimum of 100% of swap (maximum 200%). The original justification for this is that you need 100% of your memory to guarantee hibernate works (bug #311299). However, for servers, we don't hibernate and often have large amounts of memory installed, i.e. 100's of GB or even TBs, and such a method wastes a lot of disk space.
I propose we do at least 100% (probably a bit more) for machines up to a certain level of memory (4GB or less.../maybe/ up to 8GB), to ensure client machines can hibernate. However after that, we set some sort of sliding scale approach that maxes out at a reasonable amount for servers. Experienced users can always override the settings if they want.
Changed in partman-auto (Ubuntu Precise): | |
status: | New → Confirmed |
Changed in partman-auto (Ubuntu Quantal): | |
status: | New → Confirmed |
Changed in partman-auto (Ubuntu Quantal): | |
importance: | Undecided → Wishlist |
tags: | added: rls-q-tracking |
tags: |
added: rls-q-notfixing removed: rls-q-tracking |
no longer affects: | partman-auto (Ubuntu Quantal) |
While I would tend to agree, I would say that default settings (e.g. 100% memory) is for the default/majority of systems. If memory on the machine is 100's of GB or even TBs, custom partman recipes should be used to tune the machine.
While currently not supported, I would like to see magic factors a keen to check_mk magic factor for disk space notifications. [0]
[0] table at http:// mathias- kettner. de/checkmk_ filesystems. html