The Initial idea for this program seems to have come from Marius Magureanu. Credit where it's due. Timothy Clark found the code with Guillaume Quintard at Github.com/gquintard/broadcaster and added some fixes. I forked the code from Github.com/timothyclarke/http-request-broadcaster and tried to fix a panic that occasionally occurred due to a race condition that only manifested itself under high load. However, after some failed attempts I decided to completely rewrite the program to get rid of all the global variables and provide more consistent locking. To make a distinction between the original code and this new incantation, I decided to call it varnish-request-broadcaster. As of version 1.0.9, this race condition seems to be finally fixed.
Broadcasts requests to multiple Varnish caches from a single entry point. The initial thought is to ease-up purging/banning across multiple Varnish cache instances. The broadcaster consists out of a web-server which will handle the distribution of requests against all configured caches.
The easiest way is probably to pull a pre-made image from Docker Hub. Alternatively, the following can also be used:
$ go install github.com/jdavid5815/[email protected]
If you want to build from source code, then I recommend using the included Dockerfiles. Use Dockerfile.musl for lean, production ready versions. For testing and debugging, Dockerfile.glibc is recommended. It's bigger and bulkier and will consume more resources due to the inclusion of the Golang race libraries, but you'll get a lot more detailed debugging information in case of a program crash due to a race condition.
$ docker build -t jdavid5815/varnish-request-broadcaster:vX.X.X -f docker/Dockerfile.musl .
$ docker build -t jdavid5815/varnish-request-broadcaster:vX.X.X -f docker/Dockerfile.glibc .
See this file as an example on how to configure your caches.
Start the app with any of the following command line args:
- port: The port under which the broadcaster is exposed. Defaults to 8088.
- goroutines: Sets the number of available goroutines which will handle the broadcast against the caches. Defaults to a number of 8, a higher number does not necesarilly imply a better performance. Can be tweaked though depending on the number of caches.
- cfg: Path to an .ini file containing configured caches. This is a required parameter.
- retries: Number of items to retry if a request fails to execute. Defaults to 1.
- enforce: If true, the response code will be set according to the first non-200 received from the Varnish nodes.
- log-file: Path to a log file. If none specified it defaults to
stdout. - enable-log: Switches logging on/off. Disabled by default.
By default, the broadcaster starts listening on the http port, however - if both crt and key options are set, it will automatically switch onto https.
- https-port: Broadcaster https listening port. If none specified it defaults to 8443.
- crt: CRT file used for HTTPS support.
- key: KEY file used for HTTPS support.
- X-Group: Name of the group to broadcast against, if not used - the broadcast will be done against all caches.
If the broadcaster receives a SIGHUP notification, it will trigger a configuration reload from disk.
Purge /something/to/purge in all caches within the [default] group:
curl -is http://localhost:8088/something/to/purge -H "X-Group: default" -X PURGE
Ban /foo in all caches within the [prod] group:
curl -is http://localhost:8088/foo -H "X-Group: prod" -X BAN
Purge everything in all your caches:
curl -is http://localhost:8088/ -X PURGE
Purge everything in all your caches over https:
curl -is https://localhost:8443/ -X PURGE --cacert server.crt
Output sample:
HTTP/1.1 200 OK
Content-Type: application/json
Date: Tue, 06 Dec 2016 11:03:07 GMT
Content-Length: 164
{
"Cache11": 200,
"Cache12": 200,
"Cache13": 501
}
Note that your VCL needs to be aware of your purging/banning intentions. See here for more cache invalidation details.