checkpoint: support lazy migration #1541
Conversation
|
This includes the same commits as #1535 as it needs them. The only new commits are the two latest commits. |
|
@rppt: FYI |
|
Updated with the review results of #1535 |
|
@avagin: Do you have any comments on this? |
|
@kolyshkin can you also take a look at this change? Thanks! |
|
@adrianreber we need a test case for this. How it will be integrated with phaul? (https://github.com/xemul/criu/tree/criu-dev/phaul) |
|
The lazy migration feature is in a criu development branch and we are going to release it in CRIU 3.4 (Sep 2017). |
|
@avagin, I was also thinking about the test case but as the container is not totally destroyed before complete memory migration my first attempts failed as it was not possible to restore the same container with the same name. I will try to change the name, that might work. Let me try to add a working test-case for lazy migration into runC. I will update this PR. About p.haul: Not that it really belongs here, but p.haul feels kind of abandoned, especially with the difficulties of integrating p.haul and all the container engines in a useful way. So I am also leaning more towards replacing the pre-copy code in runC with CRIU's go migration library. So the main question is p.haul in its current form still alive or not? For me it seems not very alive in its current form. |
|
@adrianreber , the py version of the p.haul (that sits in a separate repo) is indeed abandoned. Mostly for the reasons you've mentioned -- too hard to integrate python code with anything else. At the same time the go p.haul, that sits in the criu repo is the place where support for criu live migration features will (well, should) go. |
|
@avagin , 3.4 is going to be in August. And lazy pages are thus aimed at 3.5. |
|
@avagin test case now included |
|
Rebased after recent breakage. Any reviewers? |
libcontainer/container_linux.go
Outdated
|
|
||
| data := make([]byte, 1) | ||
| count, _ := r.Read(data) | ||
| logrus.Debugf("%d:%s", count, status) |
There was a problem hiding this comment.
can you remove these debug statements?
libcontainer/container_linux.go
Outdated
| func waitForCriuLazyServer(r *os.File, status string) error { | ||
|
|
||
| data := make([]byte, 1) | ||
| count, _ := r.Read(data) |
There was a problem hiding this comment.
Can you add error handling for various calls in this function? Read, OpenFile, and Write are all missing
Before adding the actual lazy migration support, this adds the feature check for lazy-pages. Right now lazy migration, which is based on userfaultd is only available in the criu-dev branch and not yet in a release. As the check does not dependent on a certain version but on a CRIU feature which can be queried it can be part of runC without a new version check depending on a feature from criu-dev. Signed-off-by: Adrian Reber <[email protected]>
With the help of userfaultfd CRIU supports lazy migration. Lazy migration means that memory pages are only transferred from the migration source to the migration destination on page fault. This enables to reduce the downtime during process or container migration to a minimum as the memory does not need to be transferred during migration. Lazy migration currently depends on userfaultfd being available on the current Linux kernel and if the used CRIU version supports lazy migration. Both dependencies can be checked by querying CRIU via RPC if the lazy migration feature is available. Using feature checking instead of version comparison enables runC to use CRIU features from the criu-dev branch. This way the user can decide if lazy migration should be available by choosing the right kernel and CRIU branch. To use lazy migration the CRIU process during dump needs to dump everything besides the memory pages and then it opens a network port waiting for remote page fault requests: # runc checkpoint httpd --lazy-pages --page-server 0.0.0.0:27 \ --status-fd /tmp/postcopy-pipe In this example CRIU will hang/wait once it has opened the network port and wait for network connection. As runC waits for CRIU to finish it will also hang until the lazy migration has finished. To know when the restore on the destination side can start the '--status-fd' parameter is used: #️ runc checkpoint --help | grep status --status-fd value criu writes \0 to this FD once lazy-pages is ready The parameter '--status-fd' is directly from CRIU and this way the process outside of runC which controls the migration knows exactly when to transfer the checkpoint (without memory pages) to the destination and that the restore can be started. On the destination side it is necessary to start CRIU in 'lazy-pages' mode like this: # criu lazy-pages --page-server --address 192.168.122.3 --port 27 \ -D checkpoint and tell runC to do a lazy restore: # runc restore -d --image-path checkpoint --work-path checkpoint \ --lazy-pages httpd If both processes on the restore side have the same working directory 'criu lazy-pages' creates a unix domain socket where it waits for requests from the actual restore. runC starts CRIU restore in lazy restore mode and talks to 'criu lazy-pages' that it wants to restore memory pages on demand. CRIU continues to restore the process and once the process is running and accesses the first non-existing memory page the 'criu lazy-pages' server will request the page from the source system. Thus all pages from the source system will be transferred to the destination system. Once all pages have been transferred runC on the source system will end and the container will have finished migration. This can also be combined with CRIU's pre-copy support. The combination of pre-copy and post-copy (lazy migration) provides the possibility to migrate containers with minimal downtimes. Some additional background about post-copy migration can be found in these articles: https://lisas.de/~adrian/?p=1253 https://lisas.de/~adrian/?p=1183 Signed-off-by: Adrian Reber <[email protected]>
The lazy-pages test case is not as straight forward as the other test cases. This is related to the fact that restoring requires a different name if restored on the same host. During 'runc checkpoint' the container is not destroyed before all memory pages have been transferred to the destination and thus the same container name cannot be used. As real world usage will rather migrate a container from one system to another than lazy migrate a container on the same host this is only problematic for this test case. Another reason is that it requires starting 'runc checkpoint' and 'criu lazy-pages' in the background as those process need to be running to start the final restore 'runc restore'. CRIU upstream is currently discussing to automatically start 'criu lazy-pages' which would simplify the lazy-pages test case a bit. The handling and checking of the background processes make the test case not the most elegant as at one point a 'sleep 2' is required to make sure that 'runc checkpoint' had time to do its thing before looking at log files. Before running the actual test criu is called in feature checking mode to make sure lazy migration is in the test case criu enabled. If not, the test is skipped. Signed-off-by: Adrian Reber <[email protected]>
|
LGTM |
1 similar comment
|
LGTM |
| # there is some basic error. If the lazy migration is ready can | ||
| # be handled by $lazy_pipe. Which probably will always be ready | ||
| # after sleeping two seconds. | ||
| sleep 2 |
There was a problem hiding this comment.
It isn't only ugly, it is probably a signal that the interface is not designed properly
There was a problem hiding this comment.
@adrianreber I think we can discuss this interface on LPC
With the help of userfaultfd CRIU supports lazy migration. Lazy
migration means that memory pages are only transferred from the
migration source to the migration destination on page fault.
This enables to reduce the downtime during process or container
migration to a minimum as the memory does not need to be transferred
during migration.
Lazy migration currently depends on userfaultfd being available on the
current Linux kernel and if the used CRIU version supports lazy
migration. Both dependencies can be checked by querying CRIU via RPC if
the lazy migration feature is available. Using feature checking instead
of version comparison enables runC to use CRIU features from the
criu-dev branch. This way the user can decide if lazy migration should
be available by choosing the right kernel and CRIU branch.
To use lazy migration the CRIU process during dump needs to dump
everything besides the memory pages and then it opens a network port
waiting for remote page fault requests:
In this example CRIU will hang/wait once it has opened the network port
and wait for network connection. As runC waits for CRIU to finish it
will also hang until the lazy migration has finished. To know when the
restore on the destination side can start the '--status-fd' parameter is
used:
The parameter '--status-fd' is directly from CRIU and this way the
process outside of runC which controls the migration knows exactly when
to transfer the checkpoint (without memory pages) to the destination and
that the restore can be started.
On the destination side it is necessary to start CRIU in 'lazy-pages'
mode like this:
and tell runC to do a lazy restore:
If both processes on the restore side have the same working directory
'criu lazy-pages' creates a unix domain socket where it waits for
requests from the actual restore. runC starts CRIU restore in lazy
restore mode and talks to 'criu lazy-pages' that it wants to restore
memory pages on demand. CRIU continues to restore the process and once
the process is running and accesses the first non-existing memory page
the 'criu lazy-pages' server will request the page from the source
system. Thus all pages from the source system will be transferred to the
destination system. Once all pages have been transferred runC on the
source system will end and the container will have finished migration.
This can also be combined with CRIU's pre-copy support. The combination
of pre-copy and post-copy (lazy migration) provides the possibility to
migrate containers with minimal downtimes.
Some additional background about post-copy migration can be found in
these articles:
https://lisas.de/~adrian/?p=1253
https://lisas.de/~adrian/?p=1183
Signed-off-by: Adrian Reber [email protected]