Join GitHub today
GitHub is home to over 31 million developers working together to host and review code, manage projects, and build software together.
Sign upHead-truncate time series files only if a certain percentage of size decrease is accomplished #1296
Comments
beorn7
added
the
enhancement
label
Jan 8, 2016
beorn7
self-assigned this
Jan 8, 2016
beorn7
closed this
in
#1304
Jan 11, 2016
This comment has been minimized.
This comment has been minimized.
lock
bot
commented
Mar 24, 2019
|
This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs. |
lock
bot
locked and limited conversation to collaborators
Mar 24, 2019
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
beorn7 commentedJan 8, 2016
Servers with a very long retention period have series files with a lot of chunks. Even if only a single chunk at the beginning of the file is removed in series maintenance, the whole file is rewritten. This causes a lot of I/O for very little effect. On spinning disk, the problem doesn't show up that much because series maintenance takes long, and therefore larger changes are batched up implicitly. On SSDs, though, series maintenance is so fast and thus performed so frequently that many of those micro truncations can happen.
Some filesystems support "punching holes", which can be used to head-truncate without rewriting the file. However, that might cause more load for the filesystem, and it's not a general solution as not all filesystems support it.
An easy solution is to only rewrite the series file if a certain percentage of the total size is head-truncated. With e.g. 10%, the overhead in wasted disk space is limited but the I/O load will be reduced quite a bit.