linux/prox: timeout fifo creation#2229
Conversation
|
Please comment on the solution here. We have this in the 0.2.x branch but I don't think it is a complete solution to the FIFO lock problem. containerd/ttrpc#3 may help here, but is really unclear why the client is abandonding these fifos. |
There was a problem hiding this comment.
Value was chosen from 0.2.x branch. We may want to have something much shorter to mitigate the responsiveness problems.
Under certain conditions in the client, the fifo for a container may not be created. A timeout has been added to this operation to ensure the shim can recover when the client fails to open the fifos. Signed-off-by: Stephen J Day <[email protected]>
29d79b5 to
9754696
Compare
Codecov Report
@@ Coverage Diff @@
## master #2229 +/- ##
=======================================
Coverage 41.04% 41.04%
=======================================
Files 66 66
Lines 7753 7753
=======================================
Hits 3182 3182
Misses 4069 4069
Partials 502 502
Continue to review full report at Codecov.
|
|
Opening for write will block if the other side never open it for read (or if the process dies after opening it but before the shim had the time to open it on its side for write). The issue is that it's hard to get a universal timeout as the load of the machine affects how long it could take. This LGTM. |
|
LGTM |
Under certain conditions in the client, the fifo for a container may not
be created. A timeout has been added to this operation to ensure the
shim can recover when the client fails to open the fifos.
Signed-off-by: Stephen J Day [email protected]
Possible fix for moby/moby#36661.