cgroupsv2.LoadManager() already performs VerifyGroupPath(), and returns
an error if the path is invalid, so this check is redundant.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
Currently the shimv2 debug is only enabled when containerd is,
specifically, on debug mode. However, it should be enabled whenever the
CRI runtime is on debug *or any other lower* mode, as in trace mode.
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
fork/exec can fail and log a warning like this in containerd's log:
failed to clean up after shim disconnected error=": fork/exec /usr/local/bin/containerd-shim-[my-shim]: no such file or directory" id=test namespace=default
Passing the bundle path on the command line allows the shim delete
command to run successfully.
Signed-off-by: Samuel Karp <me@samuelkarp.com>
The shim.SetScore() utility was no longer used since 7dfc605fc6.
Checking for uses outside of this repository, I found only one external use of
this in gVisor; a9441aea27/pkg/shim/service.go (L262-L264)
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
After #4906, containerd opens fifo in read/write mode in linux platform
The original comment doesn't correct and is removed by #5174.
```
// original comment
// When using a multi-container shim, the fifo of the 2nd to Nth
// container will not be opened when the ctx is done. This will
// cause an ErrReadClosed that can be ignored.
```
However, we should add comment for checkCopyShimLogError to mention why
we call checkCopyShimLogError. The checkCopyShimLogError, it is to prevent
the flood of expected error messages after task die and the expected
errors depend on platform.
Signed-off-by: Wei Fu <fuweid89@gmail.com>
- process.Init#io could be nil
- Make sure CreateTaskRequest#Options is not empty before unmarshaling
Signed-off-by: Kazuyoshi Kato <katokazu@amazon.com>
When runC shimv2 starts, the StartShim interface will re-exec itself as
long-running process, which will read the `address` during initializing.
```happycase
Process
containerd-shim-runc-v1/v2 start containerd-shim-runc-v1/v2
initializing socket
reexec containerd-shim-runc-v1/v2
write address into file
initializing
read address
write back to containerd daemon
serving
...
remove address in Shutdown call
```
However, there is no synchronization after reexec. Then the data race is
like:
```leaking-case
Process
containerd-shim-runc-v1/v2 start containerd-shim-runc-v1/v2
initializing socket
reexec containerd-shim-runc-v1/v2
initializing
read address
write address into file
write back to containerd daemon
serving
...
fail to remove address
because of empty address
```
The `address` should be writen into file first before reexec.
And if shutdown the whole service before cleanup temporary
resource (like socket file), the Shutdown caller will receive `ttrpc: closed`
sometime, which depends on go runtime scheduler. Then it also causes leaking
socket files.
Since the shimV2-Delete binary API must be called to cleanup shim temporary
resource and shimV2-runC-v1 doesn't support grouping multi containers in one,
it is safe to remove the socket file in the binary call for shimV2-runC-v1.
But for the shimV2-runC-v2 shim, we still cleanup socket in Shutdown.
Hopefully we can find a way to cleanup socket in shimV2-Delete binary
call.
Fix: #5173
Signed-off-by: Wei Fu <fuweid89@gmail.com>
Previously a typo was introduced that caused the wrong error to be
checked against when calling exec.LookPath. This had the effect that
containerd would never locate the shim binary if it was in the same
directory as containerd's binary, but not in PATH.
Signed-off-by: Kevin Parsons <kevpar@microsoft.com>
The current code simply ignores the full binary path when starting the
shimv2 process, and instead fallbacks to a binary in the path, and this
is problematic (and confusing) for those using CRI-O, which has this
bits vendored.
The reason it's problematic with CRI-O is because the user can simply
set the full binary path and, instead of having that executed, CRI-O
will simply fail to create the container unless that binary is part of
the path, which may not be case in a few different scenarios (testing
being the most common one).
Fixes: #5006
Signed-off-by: Fabiano Fidêncio <fidencio@redhat.com>
bump version 1.3.2 for gogo/protobuf due to CVE-2021-3121 discovered
in gogo/protobuf version 1.3.1, CVE has been fixed in 1.3.2
Signed-off-by: Aditi Sharma <adi.sky17@gmail.com>
When the shim returns a plain error when a process does not exist,
the server is unable to recognise its GRPC status code and assumes
UnknownError. This is awkward for containerd client users as they are
unable to recognise the actual reason for the error.
When the shim returns a NotFound GRPC error, it is properly translated
by the server and clients receive a proper NotFound error instead of
Unknown
Please note that we (CF Garden) would like to have the eventual fix backported to 1.4 as well.
Co-authored-by: Danail Branekov <danailster@gmail.com>
Signed-off-by: Danail Branekov <danailster@gmail.com>
Signed-off-by: Georgi Sabev <georgethebeatle@gmail.com>
oom_score_adj must be in the range -1000 to 1000. In AdjustOOMScore if containerd's score is already at the maximum value we should set that value for the shim instead of trying to set 1001 which is invalid.
Signed-off-by: Simon Kaegi <simon_kaegi@ca.ibm.com>
If the shim has been killed and ttrpc connection has been
closed, the shimErr will not be nil. For this case, the event
subscriber, like moby/moby, might have received the exit or delete
events. Just in case, we should allow ttrpc-callback-on-close to
send the exit and delete events again. And the exit status will
depend on result of shimV2.Delete.
If not, the shim has been delivered the exit and delete events.
So we should remove the task record and prevent duplicate events from
ttrpc-callback-on-close.
Fix: #4769
Signed-off-by: Wei Fu <fuweid89@gmail.com>
This allows filesystem based ACLs for configuring access to the socket of a
shim.
Co-authored-by: Samuel Karp <skarp@amazon.com>
Signed-off-by: Samuel Karp <skarp@amazon.com>
Signed-off-by: Michael Crosby <michael@thepasture.io>
Signed-off-by: Michael Crosby <michael.crosby@apple.com>
The shim delete action needs bundle information to cleanup resources
created by shim. If the cleanup dead shim is called after delete bundle,
the part of resources maybe leaky.
The ttrpc client UserOnCloseWait() can make sure that resources are
cleanup before delete bundle, which synchronizes task deletion and
cleanup deadshim. It might slow down the task deletion, but it can make
sure that resources can be cleanup and avoid EBUSY umount case. For
example, the sandbox container like Kata/Firecracker might have mount
points over the rootfs. If containerd handles task deletion and cleanup
deadshim parallelly, the task deletion will meet EBUSY during umount and
fail to cleanup bundle, which makes case worse.
And also update cleanupAfterDeadshim, which makes sure that
cleanupAfterDeadshim must be called after shim disconnected. In some
case, shim fails to call runc-create for some reason, but the runc-create
already makes runc-init into ready state. If containerd doesn't call shim
deletion, the runc-init process will be leaky and hold the cgroup, which
makes pod terminating :(.
Signed-off-by: Wei Fu <fuweid89@gmail.com>
Currently the shims only support starting the logging binary process if the
io.Creator Config does not specify Terminal: true. This means that the program
using containerd will only be able to specify FIFO io when Terminal: true,
rather than allowing the shim to fork the logging binary process. Hence,
containerd consumers face an inconsistent behavior regarding logging binary
management depending on the Terminal option.
Allowing the shim to fork the logging binary process will introduce consistency
between the running container and the logging process. Otherwise, the logging
process may die if its parent process dies whereas the container will keep
running, resulting in the loss of container logs.
Signed-off-by: Akshat Kumar <kshtku@amazon.com>
Before this change, if an event fails to send on the first attempt,
subsequent attempts will fail with context.Cancelled because the the
caller of publish passes a cancellable timeout, which the publisher uses
to send the event.
The publisher returns immediately if the send fails, but adds the event
to an async queue to try again.
Meanwhile the caller will return cancelling the context.
Additionally, subsequent attempts may fail to send because the timeout
was expected to be for a single request but the queue sleeps for
`attempt*time.Second`.
In the shim service, the timeout was set to 5s, which means the send
will fail with context.DeadlineExceeded before it reaches `maxRequeue`
(which is currently 5).
This change moves the timeout to the publisher so each send attempt gets
its own timeout.
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
Previously shim v2 (`io.containerd.runc.{v1,v2}`) always used `/run/containerd/runc` as the runc root.
Fix#4326
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
How to test (from https://github.com/opencontainers/runc/pull/2352#issuecomment-620834524):
(host)$ sudo swapoff -a
(host)$ sudo ctr run -t --rm --memory-limit $((1024*1024*32)) docker.io/library/alpine:latest foo
(container)$ sh -c 'VAR=$(seq 1 100000000)'
An event `/tasks/oom {"container_id":"foo"}` will be displayed in `ctr events`.
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
Dependencies may be switching to use the new `%w` formatting
option to wrap errors; switching to use `errors.Is()` makes
sure that we are still able to unwrap the error and detect the
underlying cause.
Signed-off-by: Sebastiaan van Stijn <github@gone.nl>
TestRuntimeWithEmptyMaxEnvProcs should restore the GoMaxProcs after
test so that the temporary change of GoMaxProcs will not impact other
case, like TestRuntimeWithNonEmptyMaxEnvProcs.
Signed-off-by: Wei Fu <fuweid89@gmail.com>
For some reason, shimv2 process doesn't exist. The ttrpc doesn't detect
the connection closed by server until delete task. For this case, we
should ignore the ttrpc.ErrClosed and let task manager handle the
cleanup.
Signed-off-by: Wei Fu <fuweid89@gmail.com>
Request came from a slack message that shims do not output their versions making
it hard for users and operators to know what version of a shim they have on the
system. This adds a `-v` flag to the shims so that users can see if a shim is
in sync with containerd or what versions of shims that they are running.
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
The Err() method should be called after the Scan() loop, not inside it.
Found by: git grep -A3 -F '.Scan()'
Signed-off-by: Kir Kolyshkin <kolyshkin@gmail.com>
Instead of having several dialer implementations, leave only one in
`pkg/dialer` and call it from `pkg/ttrpcutil`, `runtime/v(1|2)/shim`
which had their own
Closes#3471.
Signed-off-by: Kiril Vladimiroff <kiril@vladimiroff.org>
The background context aovids shim blocking when the ctx is cancelled
unexpectedly during shim start. But if the shim exits unexpectedly
before opening the pipe, the fd will never be closed.
`onCloseWithShimLog` makes sure that the shim log fd is closed properly
once the shim disconnects.
Signed-off-by: Li Yuxuan <liyuxuan04@baidu.com>
There's no OOM monitoring for the v2 cgroups yet, so it seems unlikely
that there was a leak in that case.
Signed-off-by: Seth Pellegrino <spellegrino@newrelic.com>
Only start watching the cgroup for OOMs when the first process starts
instead of on every process.
Signed-off-by: Seth Pellegrino <spellegrino@newrelic.com>
If the context is cancelled during `shim.Create()`, such as the client
disconnects unexpectedly. The created shim will never be deleted.
What's more, if the context is cancelled during `openShimLog()`, the
fifo will be closed and block the shim output.
Signed-off-by: Li Yuxuan <liyuxuan04@baidu.com>
Previously, the platform was closed as part of the Delete method when the
process was an init for a task and there were no more tasks after its deletion.
This can create problems if another task is created within the shim right after
the delete runs, which results in the platform being closed but the shim
continuing to run.
This change moves closing the platform to the Shutdown method after the shim's
context is canceled, which ensures the platform is only closed once the shim
is sure its done servicing containers.
Signed-off-by: Erik Sipsma <sipsma@amazon.com>
* only shim v2 runc v2 ("io.containerd.runc.v2") is supported
* only PID metrics is implemented. Others should be implemented in separate PRs.
* lots of code duplication in v1 metrics and v2 metrics. Dedupe should be separate PR.
Signed-off-by: Akihiro Suda <akihiro.suda.cz@hco.ntt.co.jp>
Reized the I/O buffers to align with the size of the kernel buffers with fifos
and move the close aspect of the console to key off of the stdin closing.
Fixes#3738
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
- Our out of tree shim would like to publish events with ttrpc. These
functions should be exposed so our shim doesn't need to reimplement
publisher logic.
Signed-off-by: Kathryn Baldauf <kabaldau@microsoft.com>
Because of the way go handles flags, passing a flag that is not defined
will cause an error. In our case, if we kept this as a flag, then
third-party shims would break when they see this new flag. To fix this,
I moved this new configuration option to an env var. We should use env
vars from here on out to avoid breaking shim compat.
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
Previously the TTRPC address was generated as "<GRPC address>.ttrpc".
This change now allows explicit configuration of the TTRPC address, with
the default still being the old format if no value is specified.
As part of this change, a new configuration section is added for TTRPC
listener options.
Signed-off-by: Kevin Parsons <kevpar@microsoft.com>
When containerd-shim does reaper, the most processes are not init
process. Since json.Decode consumes more CPU resource, we should check
killall option for init process only.
Signed-off-by: Wei Fu <fuweid89@gmail.com>
When using a multi-container shim, the fifo of the 2nd to Nth container
will not be opened when the ctx is done. This will cause an
`ErrReadClosed` that can be ignored.
Signed-off-by: Li Yuxuan <liyuxuan04@baidu.com>
This adds a singleton `timeout` package that will allow services and user
to configure timeouts in the daemon. When a service wants to use a
timeout, it should declare a const and register it's default value
inside an `init()` function for that package. When the default config
is generated, we can use the `timeout` package to provide the available
timeout keys so that a user knows that they can configure.
These show up in the config as follows:
```toml
[timeouts]
"io.containerd.timeout.shim.cleanup" = 5
"io.containerd.timeout.shim.load" = 5
"io.containerd.timeout.shim.shutdown" = 3
"io.containerd.timeout.task.state" = 2
```
Timeouts in the config are specified in seconds.
Timeouts are very hard to get right and giving this power to the user to
configure things is a huge improvement. Machines can be faster and
slower and depending on the CPU or load of the machine, a timeout may
need to be adjusted.
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
AnonDialer will now return a "not found" error if the pipe is not found
before the timeout is reached. If the pipe exists but the timeout is
reached while attempting to connect, the timeout error will still be
returned.
This will allow the error handling logic to work properly when
connecting to the shim log pipe. An error message is only logged if the
error is not "not found", so now log noise from log pipes that were
never intended to be created by the shim will be hidden.
This change also cleans up the control flow for AnonDialer on Windows.
The new code should be more easily readable, but the only semantic
change is the error return value change.
Signed-off-by: Kevin Parsons <kevpar@microsoft.com>
When shimv2 dead, the container would be cleanup, but
the corresponding runtime task still existed in runtime
task lists, it should be deleted too.
Signed-off-by: lifupan <lifupan@gmail.com>
This changes the shim's OOM score from a static max killable of -999 to
be +1 of the containerd daemon's score. This should allow the shim's to
be killed first in an OOM condition but leave the daemon alone for a bit
to help cleanup and manage the containers during this situation.
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
This ensures that a container does not have a mounted rootfs in the
bundle directory before RemoveAll is called. Having the rootfs removed
first with a Remove ensures that the directory is not mounted and empty
before the bundle directory is removed.
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
skip hidden directories in load task, and return soon if path not exist
in atomicDelete
carry of #3233Closes#3233
Signed-off-by: Ace-Tang <aceapril@126.com>
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
Open shim v2 log with the flag `O_RDWR` will cause the `Read()` block
forever even if the pipe has been closed on the shim side. Then the
`io.Copy()` would never return and lead to a fd leak.
Fix typo when closing shim v1 log which causes the `stdouLog` leak.
Update `numPipes` function in test case to get the opened FIFO
correctly.
Signed-off-by: Li Yuxuan <liyuxuan04@baidu.com>
Before this change, the v2 runtime shim setup code was hardcoded to always
configure logrus to write logs to the "log" FIFO present in the current working
directory. This only happens in the "default" action codepath
(i.e. not shim start or shim delete).
This is problematic for shims that execute outside the current working
directory of a bundle. For example, it often doesn't make sense for shims that
manage multiple containers to execute in a single bundle directory. Additionally,
shim processes that require being pre-created, i.e. spun up before tasks they
will handle are actually created, won't have a log FIFO to write to until a task
is created.
This change leaves the default behavior as is but introduces a Binary Config
field that will optionally disable automatic configuration of logrus to use the
"log" FIFO. This allows shims to configure their own logger if necessary while
still re-using the rest of the shim helper code in containerd.
Signed-off-by: Erik Sipsma <sipsma@amazon.com>
Shims no longer call `os.Exit` but close the context on shutdown so that
events and other resources have hit the `defer`s.
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
* Rootfs dir is created during container creation not during bundle
creation
* Add support for v2
* UnmountAll is a no-op when the path to unmount (i.e. the rootfs dir)
does not exist or is invalid
Co-authored-by: Danail Branekov <danailster@gmail.com>
Signed-off-by: Georgi Sabev <georgethebeatle@gmail.com>
Use sha256 hash to shorten the unix socket path to satisfy the
length limitation of abstract socket path
This commit also backports the feature storing address path to
a file from v2 to keep compatibility
Fixes#3032
Signed-off-by: Eric Lin <linxiulei@gmail.com>
Closes#603
This adds logging facilities at the shim level to provide minimal I/O
overhead and pluggable logging options. Log handling is done within the
shim so that all I/O, cpu, and memory can be charged to the container.
A sample logging driver setting up logging for a container the systemd
journal looks like this:
```go
package main
import (
"bufio"
"context"
"fmt"
"io"
"sync"
"github.com/containerd/containerd/runtime/v2/logging"
"github.com/coreos/go-systemd/journal"
)
func main() {
logging.Run(log)
}
func log(ctx context.Context, config *logging.Config, ready func() error) error {
// construct any log metadata for the container
vars := map[string]string{
"SYSLOG_IDENTIFIER": fmt.Sprintf("%s:%s", config.Namespace, config.ID),
}
var wg sync.WaitGroup
wg.Add(2)
// forward both stdout and stderr to the journal
go copy(&wg, config.Stdout, journal.PriInfo, vars)
go copy(&wg, config.Stderr, journal.PriErr, vars)
// signal that we are ready and setup for the container to be started
if err := ready(); err != nil {
return err
}
wg.Wait()
return nil
}
func copy(wg *sync.WaitGroup, r io.Reader, pri journal.Priority, vars map[string]string) {
defer wg.Done()
s := bufio.NewScanner(r)
for s.Scan() {
if s.Err() != nil {
return
}
journal.Send(s.Text(), pri, vars)
}
}
```
A `logging` package has been created to assist log developers create
logging plugins for containerd.
This uses a URI based approach for logging drivers that can be expanded
in the future.
Supported URI scheme's are:
* binary
* fifo
* file
You can pass the log url via ctr on the command line:
```bash
> ctr run --rm --runtime io.containerd.runc.v2 --log-uri binary://shim-journald docker.io/library/redis:alpine redis
```
```bash
> journalctl -f -t default:redis
-- Logs begin at Tue 2018-12-11 16:29:51 EST. --
Mar 08 16:08:22 deathstar default:redis[120760]: 1:C 08 Mar 2019 21:08:22.703 # Warning: no config file specified, using the default config. In order to specify a config file use redis-server /path/to/redis.conf
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.704 # You requested maxclients of 10000 requiring at least 10032 max file descriptors.
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.704 # Server can't set maximum open files to 10032 because of OS error: Operation not permitted.
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.704 # Current maximum open files is 1024. maxclients has been reduced to 992 to compensate for low ulimit. If you need higher maxclients increase 'ulimit -n'.
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.705 * Running mode=standalone, port=6379.
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.705 # WARNING: The TCP backlog setting of 511 cannot be enforced because /proc/sys/net/core/somaxconn is set to the lower value of 128.
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.705 # Server initialized
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.705 # WARNING overcommit_memory is set to 0! Background save may fail under low memory condition. To fix this issue add 'vm.overcommit_memory = 1' to /etc/sysctl.conf and then reboot or run the command 'sysctl vm.overcommit_memory=1' for this to take effect.
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.705 # WARNING you have Transparent Huge Pages (THP) support enabled in your kernel. This will create latency and memory usage issues with Redis. To fix this issue run the command 'echo never > /sys/kernel/mm/transparent_hugepage/enabled' as root, and add it to your /etc/rc.local in order to retain the setting after a reboot. Redis must be restarted after THP is disabled.
Mar 08 16:08:22 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:22.705 * Ready to accept connections
Mar 08 16:08:50 deathstar default:redis[120760]: 1:signal-handler (1552079330) Received SIGINT scheduling shutdown...
Mar 08 16:08:50 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:50.405 # User requested shutdown...
Mar 08 16:08:50 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:50.406 * Saving the final RDB snapshot before exiting.
Mar 08 16:08:50 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:50.452 * DB saved on disk
Mar 08 16:08:50 deathstar default:redis[120760]: 1:M 08 Mar 2019 21:08:50.453 # Redis is now ready to exit, bye bye...
```
The following client side Opts are added:
```go
// LogURI provides the raw logging URI
func LogURI(uri *url.URL) Creator { }
// BinaryIO forwards contianer STDOUT|STDERR directly to a logging binary
func BinaryIO(binary string, args map[string]string) Creator {}
```
Signed-off-by: Michael Crosby <crosbymichael@gmail.com>
