Comment 15 for bug 1839237

Unfortunately I cannot reproduce this behavior here.

There is obviously a measurable IO when snap is refreshed, but that is expected, since a new revision needs to be downloaded (writes), downloads could be retried (read to hash the files and then writes), followed by the bookkeeping tasks (read & write). More snaps will obviously lead to the higher IO though a longer time. During the downloads, one can easily naturally expect the IO write to be as much as the download speed, since the data has to end up on disk at one point.

Would you mind helping us debugging the problem? I think it would be useful to collect the output of: `sudo pidstat -d -T ALL -p `pidof snapd` 5` and `snap changes`.

In my system I get the following stats during idle:

11:12:04 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command
11:12:05 0 4032 0.00 0.00 0.00 0 snapd

Then when for instance installing lxd:

>>>>> download still in progress
11:13:48 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command
11:13:53 0 4032 0.00 1043.20 0.00 0 snapd

11:13:53 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command
11:13:58 0 4032 5.60 925.60 0.00 0 snapd

>>>>> download done, started bookkeeping tasks
11:13:58 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command
11:14:03 0 4032 1048.80 1677.60 138.40 0 snapd

>>>>> all done, back to idle
11:14:03 UID PID kB_rd/s kB_wr/s kB_ccwr/s iodelay Command
11:14:08 0 4032 0.00 0.00 0.00 0 snapd