This change updates how we handle long lines of output from the
container. The previous logic used a bufio reader to read entire lines
of output from the container through an intermediate BytesPipe, and that
allowed the container to cause dockerd to consume an unconstrained
amount of memory as it attempted to collect a whole line of output, by
outputting data without newlines.
To avoid that, we replace the bufio reader with our own buffering scheme
that handles log lines up to 16k in length, breaking up anything longer
than that into multiple chunks. If we can dispense with noting this
detail properly at the end of output, we can switch from using
ReadBytes() to using ReadLine() instead. We add a field ("Partial") to
the log message structure to flag when we pass data to the log driver
that did not end with a newline.
The Line member of Message structures that we pass to log drivers is now
a slice into data which can be overwritten between calls to the log
driver's Log() method, so drivers which batch up Messages before
processing them need to take additional care: we add a function
(logger.CopyMessage()) that can be used to create a deep copy of a
Message structure, and modify the awslogs driver to use it.
We update the jsonfile log driver to append a "\n" to the data that it
logs to disk only when the Partial flag is false (it previously did so
unconditionally), to make its "logs" output correctly reproduce the data
as we received it.
Likewise, we modify the journald log driver to add a data field with
value CONTAINER_PARTIAL_MESSAGE=true to entries when the Partial flag is
true, and update its "logs" reader to refrain from appending a "\n" to
the data that it retrieves if it does not see this field/value pair (it
also previously did this unconditionally).
Signed-off-by: Nalin Dahyabhai <nalin@redhat.com> (github: nalind)
inotify event is trigged immediately there's data written to disk.
But at the time that the inotify event is received, the json line might
not fully saved to disk. If the json decoder tries to decode in such
case, an io.UnexpectedEOF will be trigged.
We used to retry for several times to mitigate the io.UnexpectedEOF error.
But there are still flaky tests caused by the partial log entries.
The daemon knows exactly when there are new log entries emitted. We can
use the pubsub package to notify all the log readers instead of inotify.
Signed-off-by: Shijiang Wei <mountkin@gmail.com>
try to fix broken test. will squash once tests pass
Signed-off-by: Shijiang Wei <mountkin@gmail.com>
- Move time json marshaling to the jsonlog package: this is a docker
internal hack that we should not promote as a library.
- Move Timestamp encoding/decoding functions to the API types: This is
only used there. It could be a standalone library but I don't this
it's worth having a separated repo for this. It could introduce more
complexity than it solves.
Signed-off-by: David Calavera <david.calavera@gmail.com>
The json decoder starts to decode immediately an inotify event is
received.
But at the time the inotify event is trigged, the json log
entry might haven't been fully written to the disk.
In this case the decoder will return an "io.UnexpectedEOF" error, but
there is still data remaining in the decoder's buffer. And the data
should be passed to the decoder when the next inotify event is
triggered.
Signed-off-by: Shijiang Wei <mountkin@gmail.com>
this allows jsonfile logger to collect extra metadata from containers with
`--log-opt labels=label1,label2 --log-opt env=env1,env2`.
Extra attributes are saved into `attrs` attributes for each log data.
Signed-off-by: Daniel Dao <dqminh@cloudflare.com>
After tailing a file, if the number of lines requested is > the number
of lines in the file, this would cause a json unmarshalling error to
occur when we later try to go follow the file.
So brute force set it to the end if any tailing occurred.
There is potential that there could be some missing log messages if logs
are being written very quickly, however I was not able to make this
happen even with `while true; do echo hello; done`, so this is probably
acceptable.
While testing this I also found a panic in LogWatcher.Close can be
called twice due to a race. Fix channel close to only close when there
has been no signal to the channel.
Signed-off-by: Brian Goff <cpuguy83@gmail.com>
- downcase and privatize exported variables that were unused
- make accurate an error message
- added package comments
- remove unused var ReadLogsNotSupported
- enable linter
- some spelling corrections
Signed-off-by: Morgan Bauer <mbauer@us.ibm.com>
- noplog driver pkg for '--log-driver=none' (null object pattern)
- centralized factory for log drivers (instead of case/switch)
- logging drivers registers themselves to factory upon import
(easy plug/unplug of drivers in daemon/logdrivers.go)
- daemon now doesn't start with an invalid log driver
- Name() method of loggers is actually now their cli names (made it useful)
- generalized Read() logic, made it unsupported except json-file (preserves
existing behavior)
Spotted some duplication code around processing of legacy json-file
format, didn't touch that and refactored in both places.
Signed-off-by: Ahmet Alp Balkan <ahmetalpbalkan@gmail.com>
