lukepalmer / brubeck Goto Github PK

We are trying to build graphite-statsd Dockerfile on s390x But our image build is failing with below error

Step 29/45 : RUN git clone "${brubeck_repo}" /usr/local/src/brubeck  && cd /usr/local/src/brubeck && git checkout "${brubeck_version}"  && ./script/bootstrap  && chmod +x brubeck && mkdir -p /opt/graphite/bin/  && cp -fv brubeck /opt/graphite/bin/brubeck
 ---> Running in bf8c5e51439f
Cloning into '/usr/local/src/brubeck'...
Note: switching to 'e95d95945a9319be38e593e95691344a99fe851d'.

You are in 'detached HEAD' state. You can look around, make experimental
changes and commit them, and you can discard any commits you make in this
state without impacting any branches by switching back to a branch.

If you want to create a new branch to retain commits you create, you may
do so (now or later) by using -c with the switch command. Example:

  git switch -c <new-branch-name>

Or undo this operation with:

  git switch -

Turn off this advice by setting config variable advice.detachedHead to false

HEAD is now at e95d959 Finish removing secure (missed some spots)
Submodule 'vendor/ck' (https://github.com/concurrencykit/ck) registered for path 'vendor/ck'
Cloning into '/usr/local/src/brubeck/vendor/ck'...
Submodule path 'vendor/ck': checked out '1a84d49c3ca794356f015a3391e6d10be98f6a6a'
cd vendor/ck && ./configure
Detecting operating system.......success [linux]
Detecting machine architecture...failed  [unsupported]
make: *** [Makefile:60: vendor/ck/Makefile] Error 1
The command '/bin/sh -c git clone "${brubeck_repo}" /usr/local/src/brubeck  && cd /usr/local/src/brubeck && git checkout "${brubeck_version}"  && ./script/bootstrap  && chmod +x brubeck && mkdir -p /opt/graphite/bin/  && cp -fv brubeck /opt/graphite/bin/brubeck' returned a non-zero code: 2

This failure is observed because checked out version of concurrencykit/ck does not have s390x support. I could build latest https://github.com/concurrencykit/ck/releases/tag/0.7.0 on s390x platform
@lukepalmer Could you please update this repo to use latest stable version(0.7.0) of concurrencykit/ck?

Outputting time in fractional milliseconds doesn't make a lot of sense in a macro context. Optionally normalize to seconds.

This should be configurable so that people who have been using milliseconds forever can keep continuity.

Got error in make
src/backends/kafka.c:40:9: warning: implicit declaration of function 'rd_kafka_fatal_error'; did you mean 'rd_kafka_last_error'? [-Wimplicit-function-declaration] err = rd_kafka_fatal_error(self->rk, errstr, sizeof(errstr)); ^~~~~~~~~~~~~~~~~~~~ rd_kafka_last_error src/backends/kafka.c: In function 'error_cb': src/backends/kafka.c:96:14: error: 'RD_KAFKA_RESP_ERR__FATAL' undeclared (first use in this function); did you mean 'RD_KAFKA_RESP_ERR__FAIL'? if (err != RD_KAFKA_RESP_ERR__FATAL) ^~~~~~~~~~~~~~~~~~~~~~~~ RD_KAFKA_RESP_ERR__FAIL src/backends/kafka.c:96:14: note: each undeclared identifier is reported only once for each function it appears in make: *** [src/backends/kafka.o] Error 1

"Client-sent sampling rates are ignored."
Is it an outdated statement in readme?

Do brubeck uses only 65536 samples for ms aggregates?

The existing implementation appears to be taken from here:
https://github.com/karelzak/util-linux/blob/master/lib/setproctitle.c

This one is probably better:
https://lwn.net/Articles/462095/

should be brubeck and not brubeck-new

For example: if you're not using statsd secure then don't output stats for it.

There are deprecation warnings in the calls made to libmicrohttpd. Update these.

Right now there's a fixed size in the config, and it's easy for performance to degrade if you don't think about it.

This could happen automatically using ck_ht_grow_spmc.

There was an update to an underlying library that did not stand the test of time. There are compilation warnings and I suspect the implementation does not work.

For the next person who needs StatsD secure (does anyone need/use this?) you may need to make some hopefully minor fixes before use. PR appreciated of course.

Right now values are guarded using pthread_spin_lock. It would be faster to use GCC atomic builtins.

Most people probably do not need a dependency on librdkafka.