comment
This commit is contained in:
parent
5da1a78508
commit
21cf33a881
1 changed files with 29 additions and 0 deletions
|
@ -0,0 +1,29 @@
|
|||
[[!comment format=mdwn
|
||||
username="joey"
|
||||
subject="""re: comment 16"""
|
||||
date="2022-06-06T16:20:55Z"
|
||||
content="""
|
||||
@Atemu, that seems like at least a plausible theory, that disk write
|
||||
pressure (eg from getting annex objects) is causing sqlite writes to take a
|
||||
long time. Increasing the wait time from the current 10 seconds would make
|
||||
sense if that's what's happening.
|
||||
|
||||
The somewhat weird thing is that, with a single git-annex process, no
|
||||
matter the number of jobs, it only has a single database handle open,
|
||||
and so it should not be possible for one write to still be happening when
|
||||
a second write happens. Unless, of course, sqlite is doing something behind
|
||||
my back, like finishing the write in a separate thread? Or, of course,
|
||||
there might be a second process that has somehow not been visible..
|
||||
|
||||
So, I've added some debugging of when it commits to the database.
|
||||
@yarik, please get an updated build with commit
|
||||
5da1a785086910c32559a46f9110779d765af0cd,
|
||||
and try after setting:
|
||||
|
||||
git config annex.debug true
|
||||
git config annex.debugfilter Database.Handle
|
||||
|
||||
This should tell us if there are somehow multiple writers at the same time,
|
||||
and also will tell us the timing of how long it is taking to commit to the
|
||||
db.
|
||||
"""]]
|
Loading…
Add table
Add a link
Reference in a new issue