Dynamic configuration gRPC service and few other enhancements and bug fixes

[NOTE: This release is deprecated because of #264. Use v0.10.4 instead.]

• Add a gRPC to dynamically configure cloudprober. You'll be able to add/remove/list probes using this service. Here is the Go API to interact with this server: CloudproberServer. Example client to work with the gRPC service: client.go.

• Extend external probe to add ability to parse "map" and "distribution" metric types from the probe output.

• Streamline the usage of source_ip_config and ip_version options throughout cloudprober. These options can now be configured at the outer layer instead of probe-type specific configs. This change makes the usage of these fields much more predictable and consistent.

All changes since the last release:
v0.10.2...v0.10.3

Performance improvements, some features and a bug fix

This release includes a bunch of performance improvements which are noticeable if you use cloudprober with many targets. HTTP probe and server get more options. A minor bug fix in the prometheus surfacer.

Performance improvements

• Heap usage performance (and hence GC performance) improvements in ping and http probes. We cleaned up hot paths to make them use less and less heap. We also improved string building throughout the codebase.
• Better distribution of probe runs across time. This improves latency measurements quite noticeably if you run a lot of probes.

Features

• HTTP probe: added support for enabling and disabling of HTTP/2.
• HTTP probe: added support for controlling the TCP keep-alive option.
• Added latency distributions support to DNS and EXTERNAL probes.
• Cloudprober's HTTP server supports HTTP/S now.

Bug fix

• Prometheus surfacer had a hard-to-reproduce bug where a panic could be triggered in the HTTP handler on concurrent access (#222).

See all the changes here:
v0.10.1...v0.10.2

v0.10.1

Bug Fix

Fixes a bug in the file surfacer when compression is enabled (see commit: 17ed0af).

v0.10.0

Validators framework

Now you can configure validators for your probes. A validator is a test that can be run on the probe response to verify its validity, e.g. a regex match, status code for HTTP response, integrity check, etc. See validators example for more information: cloudprober_validator.cfg.

Status page

Cloudprober now provides a status page at the "/status" URL. Status page shows cloudprober version, uptime, and currently configured probes, surfacers and servers.

Other changes

• Add a new server type - gRPC.
• Add support for writing compressed data through FILE surfacer (9382a5a). This will help scaling cloudprober further for lots of targets and probes.
• Add distributions support to PING probes (Issue #188).
• Add support for specifying source IP for HTTP probes.
• Add support for exponential distributions.
• Add documentation on extending cloudprober.

Enhancements and bug fixes

I'll update the release notes shortly, but there have been quite a few enhancements and bug fixes since the last release.

Enhancements, bug fixes

Couple of exciting new features and a couple of bug fixes.

• Support for distributions (histograms). Cloudprober now supports distribution/histogram metrics. Latency metric ('latency') can now be configured to be a distribution. External probes can also export their metrics as distributions. More documentation on how to use them to be added soon. See issue #13 for more details.

• UDP probe type redone to be more efficient. It now uses a fixed set of sockets for multiple targets. Number of sockets can be pre-configured (#10).

• Couple of performance bug fixes:

  • Map variable's stringer (String()) was inefficient. If a large map was created, for example by accessing a lot of URLs on cloudprober's HTTP server, exporting map variable as a string will be expensive.
  • External probe type (in SERVER mode) didn't handle external probe process death very cleanly.

Bug fix release

• Contains the fix for timeout issue in external probe in 'SERVER' mode (#29).

• Switch to non-alpha APIs for GCE so that cloudprober doesn't require alpha API access.

Few enhancements

Few metrics standardization enhancements:

• All probes export the same set of metrics: total/success/latency (replaced sent/rcvd/rtt).
• External probe generated metrics can be CUMULATIVE or GAUGE, based on the config.

Improved GCE instances refresh logic:

• Only use 'RUNNING' instances.
• Introduce a random delay before starting the refresh cycle => If there are multiple cloudprober instances running, all of them will not refresh instance list at the same time.

Cloudprober beta release

You can always download the latest docker image from docker hub:
docker run cloudprober/cloudprober

To build and install from the source:
go get github.com/google/cloudprober
GOBIN=$GOPATH/bin go install $GOPATH/src/github.com/google/cloudprober/cmd/cloudprober.go

0.0.1

First cloudprober release