Diminishing performance on large files
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Duplicity |
Fix Released
|
Medium
|
Unassigned | ||
0.6 |
Fix Released
|
Medium
|
Unassigned |
Bug Description
I am testing duplicity for backing up my VMs. Using version 0.6.08b, it took 14 hours to transfer a 28.8 GB VM to Amazon's S3 on my FiOS connection, doing a FULL backup. I have sustained well over 2 MB/s on our business line for over 1 hour, so I'm sure the delay/lag is coming from duplicity. The server I am running on is a brand new box with 4GB RAM, 500GB of space, core i5 quad core and is freshly installed with ARCH linux for testing duplicity and nothing more.
I then tried again, but applied this patch here: https:/
I am using the --asynchronous-
I am using Python 2.6 on Arch x86-64 (Fully updated and freshly installed).
Changed in duplicity: | |
importance: | Undecided → Medium |
assignee: | nobody → Kenneth Loafman (kenneth-loafman) |
milestone: | none → 0.6.09 |
Changed in duplicity: | |
status: | New → In Progress |
Changed in duplicity: | |
status: | In Progress → Incomplete |
no longer affects: | duplicity/0.7-series |
Changed in duplicity: | |
status: | In Progress → Fix Released |
I tried restarting the upload (fresh, deleting my s3 bucket) and the drop in performance came MUCH quicker, only after 10 minutes while i was monitroing it. I've attached the last 70 lines or so of a -v9 output. It keeps hanging at "AsyncScheduler: scheduling task for asynchronous execution" for about a minute before continuing (started at volume 43).
I also notice that the creation of the next volume waits before the previous one was done uploading. It won't say "Proccessed <next volume>" until after the upload is finished. Not sure if that's just coincidental.