mirror/runc
1
0
Fork 0
mirror of https://github.com/opencontainers/runc.git synced 2026-07-11 06:03:57 +08:00
Code Issues Packages Projects Releases Wiki Activity

tests/int/checkpoint: fix lazy migration flakiness

Browse Source 
When doing a lazy checkpoint/restore, we should not restore into the
same cgroup, otherwise there is a race which result in occasional
killing of the restored container (GH #2760, #2924).

The fix is to use --manage-cgroup-mode=ignore, which allows to restore
into a different cgroup.

Note that since cgroupsPath is not set in config.json, the cgroup is
derived from the container name, so calling set_cgroups_path is not
needed.

For the previous (unsuccessful) attempt to fix this, as well as detailed
(and apparently correct) analysis, see commit 36fe3cc28c.

Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
This commit is contained in:
Kir Kolyshkin
2022-08-04 11:57:31 -07:00
parent 683528783e
commit c4aa452b18
1 changed files with 18 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/integration/checkpoint.bats
+18 -7
View File
@@ -224,7 +224,14 @@ function simple_cr() {
# TCP port for lazy migration
port=27277
__runc checkpoint --lazy-pages --page-server 0.0.0.0:${port} --status-fd ${lazy_w} --work-path ./work-dir --image-path ./image-dir test_busybox &
__runc checkpoint \
--lazy-pages \
--page-server 0.0.0.0:${port} \
--status-fd ${lazy_w} \
--manage-cgroups-mode=ignore \
--work-path ./work-dir \
--image-path ./image-dir \
test_busybox &
cpt_pid=$!
# wait for lazy page server to be ready
@@ -246,14 +253,18 @@ function simple_cr() {
lp_pid=$!
# Restore lazily from checkpoint.
# The restored container needs a different name (as well as systemd
# unit name, in case systemd cgroup driver is used) as the checkpointed
# container is not yet destroyed. It is only destroyed at that point
# in time when the last page is lazily transferred to the destination.
#
# The restored container needs a different name and a different cgroup
# (and a different systemd unit name, in case systemd cgroup driver is
# used) as the checkpointed container is not yet destroyed. It is only
# destroyed at that point in time when the last page is lazily
# transferred to the destination.
#
# Killing the CRIU on the checkpoint side will let the container
# continue to run if the migration failed at some point.
[ -v RUNC_USE_SYSTEMD ] && set_cgroups_path
runc_restore_with_pipes ./image-dir test_busybox_restore --lazy-pages
runc_restore_with_pipes ./image-dir test_busybox_restore \
--lazy-pages \
--manage-cgroups-mode=ignore
wait $cpt_pid