prefix has been removed as a field in Cortex and needs to be removed
from the ksonnet as well.

* update cortex to latest master

using v3.4.0-rc.1 for go.etcd.io/etcd
v0.0.2 for github.com/prometheus/procfs
v1.0.0 for github.com/prometheus/client_golang
latest master for github.com/weaveworks/common with changes from https://github.com/weaveworks/common/pull/153

* fixed failing tests

* use large instance for test and lint jobs in CircleCI

* running only 6 test binaries in parallel

* removed resource type change for CircleCI

* changed GOGC to 10 for lint makefile target

* Do not send the last batch from promtail on graceful shutdown if the batch is empty

* Tested promtail Client.Handle()

The journal scrape config in promtail has been updated to support
a "max_age" field. max_age determines the oldest journal entry
promtail will read when starting the journal reader. When
unspecified, max_age defaults to 7h.

Even if a position in the journal is saved in the promtail
positions file, that position may be ignored if the entry corresponding
to that position is older than the max_age cutoff time.

Example promtail config for getting up to the last 12h of journal
entries:

  server:
    http_listen_port: 9080
    grpc_listen_port: 0

  positions:
    filename: /tmp/positions.yaml

  clients:
    - url: http://localhost:3100/api/prom/push

  scrape_configs:
  - job_name: journal
    journal:
      max_age: 12h
      path: /var/log/journal
      labels:
        job: systemd-journal
    relabel_configs:
      - source_labels: ['__journal__systemd_unit']
        target_label: 'unit'

* add quota for sed

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

* add comment for head

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

adding the beginnings of the release doc with steps, fixing a bug in the release_prepare.sh script and tweaking the output based on the current process.

* improvements in bigtable backup tool

Added some metrics for monitoring of backup jobs
Added a command to delete out of range backups, otherwise, there would be more backups than expected and alerts would start firing for unexpected backups

* Updating changelog for v0.3.0

* Fixing consecutive list problem by putting an empty html comment between lists

* chore(packaging): set tag length to 7

relied on git approximation before, which causes problems with other parts of
the automation.

Because we do not need the tag to be strictly unique, but just to identify the
latest version, limiting to 7 is fine.

If this ever causes problems, raise the limit but make sure to update all other
places as well.

* chore(packaging): request at least 7 characters

Co-Authored-By: Dan Cech <dan@aussiedan.com>

brrrrrrr

Because drone.io provides us with native ARM runners, we evaluate the platform to figure out if it is a suitable replacement for CircleCI and qemu.

Configuration is done using .drone/drone.jsonnet, the .drone/drone.yml is generated from that.
Images are built on amd64, arm64 and arm, natively and without emulation. This allows to get rid of cross-compilation, which in turn allows more advanced builds (CGO, etc.)

Hopefully this will make the CI faster and more reliable in the future. Let's see!

ingester.max_transfer_retries and
ingester.lifecycler.{join_after, claim_on_rollout} have been
added and set to match values proven to work for Cortex in production.

Shutting the module down before shutting down the HTTP server enables
metrics to continue to be collected while the module terminates.

This change syncs up with how Cortex modules are shut down; the module
is always terminated before the HTTP server.

Fixes #819.

Fix link format

feat(ksonnet): make our ksonnet a little more modular by parameterizing the chunk and index stores (#874)

* make the missing entry lookup synchronous during tests to hopefully avoid races around the entry order and test verifications

* playing around with timings some more to get all tests to pass even on a really slow machine

File permissions on creation of positions.yaml will now be 0600
instead of 0700.

Fixes #892.

`from: 0` throws
```
level=error ts=2019-08-07T12:02:15.385981107Z caller=main.go:48 msg="error loading config" filename=/etc/loki/local-config.yaml err="parsing time \"0\" as \"2006-01-02\": cannot parse \"0\" as \"2006\""
```

`store: dynamo` throws:
```
level=error ts=2019-08-07T11:59:01.688427565Z caller=main.go:71 msg="error initialising loki" err="error initialising module: table-manager: Unrecognized storage client dynamo, choose one of: aws, cassandra, inmemory, gcp, bigtable, bigtable-hashed"
```

This commit introduces a stopgap to get ARM builds working by
temporarily disabling journal reading support on ARM platforms.

Due to an invalid usage of bash, the binary was always built for amd64.
This fixes that and builds the correct one.

This commit modifies the Makefile to build promtail with CGO_ENABLED if
both GOHOSTOS and GOOS are set to linux.

To facilitate the Docker container utilizing the journal support,
the container for promtail had to be modified to be based on top of
debian. The sd-journal library depends on libsystemd.so being present
at runtime. As alpine currently does not have a package for systemd,
the easiest approach is to change the base image to debian:stretch-slim.
A future solution for providing an alpine-based build will be considered
in the future and discussion for the approach will be tracked in #838.

This commit also updates the Dockerfile to default to using the
latest loki-build-image. If a specific version is required, it should
be provided as a build arg to docker build.

update retention section in docs to recommend retention to be twice the duration of periodic table

* Fixed labels process test with same objects

Signed-off-by: jojohappy <sarahdj0917@gmail.com>

Removing the loki-frontend dashboard as it only has one panel which is duplicated in the loki-reads panel, we can bring it back when add more to the loki frontend.