* Adding a sync() function to filetarget which runs on a timer to catch any events missed by fsnotify

* moving where we mark the final position to avoid a race condition exposed in testing, we want to stop the tailer before shutting down our tailer loop

* Fix changes from PR and refactor to remove circular imports and make config consistent

* Adding an "integration" type unit test which tests promtail as a whole

* fixing a bunch of imports I munged
moved tailer to its own file
refactoring filetarget to make synclogic easier to follow and reused as much as possible

* update helm values yaml

- set persistence.enabled as 'false' as default since there is no pv as
default

- set persistence.storageClassName as 'default' since even if pv exists, pvc
still need storageClassName:

```
  Normal  FailedBinding  8s (x4 over 35s)  persistentvolume-controller  no persistent volumes available for this claim and no storage class is set
```

Signed-off-by: Xiang Dai <764524258@qq.com>

* helm: delete duplicate empty blank

Signed-off-by: Xiang Dai <764524258@qq.com>

Added info about GrafanaCon at the top

dont add pods twice if they have app and name labels

issue #313

Add k8s part in docs

* Include the stream's labels in OOO error responses.

Also, read the body of error responses and log them.  And retries 500s and connection errors.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

* Log close errors.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

* Update pkg/promtail/client/client.go

Co-Authored-By: tomwilkie <tomwilkie@users.noreply.github.com>

* Log retries at warn, final errors at error.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

* set currLabels when creating new heapIterator

fixes #307

* update unit tests to ensure labels are set correctly

Proves that they can accept multiple samples with the same timestamp.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Co-Authored-By: yubozhao <yubz86@gmail.com>

Signed-off-by: Julien Garcia Gonzalez <julien@giantswarm.io>

This was broken in a422f394.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

alpine:3.4 only contains a docker image for arm64, so building for other
architectures doesn't work. The newest alpine relases, 3.9, is therefore
better.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

fix promtail_request_duration metric

I broke this when I updated the vendored Cortex, which includes https://github.com/cortexproject/cortex/commit/ac653065193fc9f9b91d0a3da38886cd20b816e5#diff-42196489d8c6339357699ab3656cb2e1R315

.

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

When starting up, start all dependencies first then start the target
module.  When stopping, reverse the order to ensure we dont try and
stop a module while there is still something running that needs it.

fixes #105

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>

Signed-off-by: Tom Wilkie <tom.wilkie@gmail.com>