virt plugin: Add support for domain tags
One "domain tag" is one custom attribute in the libvirt domain
metadata section. This patch enhances the virt plugin to partition
the domains to sample into reader instances according to tags.
One reader instance will only query the domains with attached
a macthing tag.
The special-purpose reader instance #0, guaranteed to be always present,
(since <0 instances are not allowed), will query all the domains
with missing or unrecognized tag, so no domain will ever left
out.
It's up to one external entity, like a management application,
to properly tag domains however it sees fit; how tags are picked
is completely transparent to the plugin
Tagging could be used by management application to evenly spread the
load among the reader threads, or to pin on the same threads all
the libvirt domains which use the same shared storage, to minimize
the disruption in presence of storage outages.
Signed-off-by: Francesco Romani <fromani@redhat.com>
One "domain tag" is one custom attribute in the libvirt domain
metadata section. This patch enhances the virt plugin to partition
the domains to sample into reader instances according to tags.
One reader instance will only query the domains with attached
a macthing tag.
The special-purpose reader instance #0, guaranteed to be always present,
(since <0 instances are not allowed), will query all the domains
with missing or unrecognized tag, so no domain will ever left
out.
It's up to one external entity, like a management application,
to properly tag domains however it sees fit; how tags are picked
is completely transparent to the plugin
Tagging could be used by management application to evenly spread the
load among the reader threads, or to pin on the same threads all
the libvirt domains which use the same shared storage, to minimize
the disruption in presence of storage outages.
Signed-off-by: Francesco Romani <fromani@redhat.com>
virt plugin: Support for multiple read callbacks
Add support to register more than one read callbacks in the
reader thread pool.
The default configuration is to use just one read callback,
for backward compatibility. No user-visible changes are expected
in this scenario.
Signed-off-by: Francesco Romani <fromani@redhat.com>
Add support to register more than one read callbacks in the
reader thread pool.
The default configuration is to use just one read callback,
for backward compatibility. No user-visible changes are expected
in this scenario.
Signed-off-by: Francesco Romani <fromani@redhat.com>
virt plugin: Factor the read state into a struct
The lv_read function needs some bookkeeping data to track which
domain, block interface and network interface should be polled
for new data.
This patch factors out this data, previously scattered as module
globals, in a new struct. This makes the code a little tidier
and more reusable.
Signed-off-by: Francesco Romani <fromani@redhat.com>
The lv_read function needs some bookkeeping data to track which
domain, block interface and network interface should be polled
for new data.
This patch factors out this data, previously scattered as module
globals, in a new struct. This makes the code a little tidier
and more reusable.
Signed-off-by: Francesco Romani <fromani@redhat.com>
Bump version to 5.6.1; Update ChangeLog.
Merge pull request #1898 from rubenk/remove-libltdl-support
Replace libltdl with plain dlopen.
Replace libltdl with plain dlopen.
contrib/docker: cleanup wrapper a bit
- No need to declare errno
- Remove one level of indentation
- No need to declare errno
- Remove one level of indentation
Drop support for libtool version 1
The only distro that we still support that uses libtool version 1
is RHEL5, but that will be EOL in a few months.
The only distro that we still support that uses libtool version 1
is RHEL5, but that will be EOL in a few months.
src/utils_format_graphite.[ch]: Test correct null termination.
Issue: #2063
Issue: #2063
Merge branch 'collectd-5.7'
Merge pull request #2064 from rubenk/fix-make-distcheck
Fix make distcheck
Fix make distcheck
Collected podtypo
Merge branch 'collectd-5.7'
Merge branch 'collectd-5.6' into collectd-5.7
Conflicts:
contrib/redhat/collectd.spec
Conflicts:
contrib/redhat/collectd.spec
collectd.spec: also disable write_redis on EL6
This plugin also depends on the hiredis library, which was retired from
EPEL6.
This plugin also depends on the hiredis library, which was retired from
EPEL6.
Merge branch 'collectd-5.5' into collectd-5.6
Conflicts:
contrib/redhat/collectd.spec
Conflicts:
contrib/redhat/collectd.spec
contrib: clang-format rootfs_prefix.c
collectd.spec: disable redis plugin on EL6
hiredis has been retired from EPEL6
hiredis has been retired from EPEL6
collectd.spec: disable redis plugin on EL6
Merge pull request #2041 from mfournier/contrib-docker
Add sample Dockerfile & LD_PRELOAD wrapper to contrib
Add sample Dockerfile & LD_PRELOAD wrapper to contrib
write_tsdb plugin: Simplify the ResolveInterval configuration.
The defaults have been set to the more reasonable 1*interval.
The defaults have been set to the more reasonable 1*interval.
write_tsdb plugin: Use cdrand_range().
Also rename the global variables to match the config options.
Also rename the global variables to match the config options.
collectd.conf(5): s/TSDN/TSDB/
write_tsdb plugin: Rename options to "Resolve{Interval,Jitter}".
write_tsdb plugin: Coding style fixes.
* Rename ->sock_info to ->ai.
* Use cf_utils_get_cdtime() to parse time options.
* Optimize away the use_dnsrandomttl global variable.
* Rename ->sock_info to ->ai.
* Use cf_utils_get_cdtime() to parse time options.
* Optimize away the use_dnsrandomttl global variable.
write_tsdb : Add a random TTL before querying the DNS again
Bugfix on write_tsdb configuration parser
Prevent DNS flood when TSDB is not available
Even more renormalization.
Merge branch 'collectd-5.6'
More reformatting.
No idea why the previous round missed these files.
No idea why the previous round missed these files.
Merge branch 'collectd-5.6'
Tree wide: Reformat with clang-format.
Tree wide: Reformat with clang-format.
Merge branch 'collectd-5.5' into collectd-5.6
Bump version to 5.5.3; Update ChangeLog.
Merge branch 'pr/2056'
collectd.conf(5): Improve documentation of the turbostat plugin.
Updated to actually honor the boolean and clarified documentation per request
collectd.conf(5): Add missing blank line.
src/utils_latency_config.c: Avoid nested compound literals.
So GCC 4.6 doesn't crash.
So GCC 4.6 doesn't crash.
src/daemon/utils_match.[ch]: Rename UTILS_MATCH_CF_GAUGE_LATENCY to UTILS_MATCH_CF_GAUGE_DIST.
tail plugin: Rename DSType from Latency to Distribution.
This is a more flexible naming owing to the fact that percentiles and other
distribution parameters are interesting for a variety of metrics, not just
latencies.
The config handling is now completely in src/utils_latency_config.c so
that other plugins, such as the cURL plugin, can easily reuse the module
with consistent config handling.
This is a more flexible naming owing to the fact that percentiles and other
distribution parameters are interesting for a variety of metrics, not just
latencies.
The config handling is now completely in src/utils_latency_config.c so
that other plugins, such as the cURL plugin, can easily reuse the module
with consistent config handling.
src/utils_latency{,_config}.[ch]: Reformat new code with clang-format.
collectd.conf(5): Update the "LatencyRate" option.
src/utils_latency.[ch]: get_rate(): Make lower bound exclusive.
This has a bunch of benefits:
* You can easily iterate over a range of latencies without counting
latencies twice. This was previously tricky because both borders were
considered to be inclusive.
* When lower equals upper, the returned value is now zero.
Previously, it was a value very close to zero, but not zero. The exact
value depended on the bucket width, an information not easily
available to the caller.
This has a bunch of benefits:
* You can easily iterate over a range of latencies without counting
latencies twice. This was previously tricky because both borders were
considered to be inclusive.
* When lower equals upper, the returned value is now zero.
Previously, it was a value very close to zero, but not zero. The exact
value depended on the bucket width, an information not easily
available to the caller.
src/utils_latency.[ch]: Remove latency_counter_get_start_time().
src/utils_latency.[ch]: Improve accuracy, update unit test.
DSType latency: Improved after PR code review
+ liblatency: Added utils_latency_config code
+ tail plugin: DSType Latency added
+ tail plugin: DSType Latency added
+ utils_latency: Cleaned latency_counter_get_rate() from debugging
+ utils_latency: Added latency_counter_get_rate()
+ User_data destroy callback added to match_create_callback() in utils_match
Merge remote-tracking branch 'github/pr/2058'
amqp, write_graphite and write_kafka plugins: Implement the "[Graphite]PreserveSeparator" option.
Fixes: #419
Fixes: #419
src/utils_format_graphite.[ch]: Implement the GRAPHITE_PRESERVE_SEPARATOR flag.
Add config boolean for logical CPU naming for turbostat plugin
Merge remote-tracking branch 'github/pr/2052'
Merge remote-tracking branch 'github/pr/2054'
src/utils_format_graphite.[ch]: Add unit test.
First step towards fixing issue #419.
First step towards fixing issue #419.
collectd.spec: enable write_prometheus
use deterministic time values in MockDaemon
recognize seconds with and without fractional part
pass fractional seconds upwards as floats
allow fractional seconds in listval_filter()
update MockDaemon to report fractional seconds
fix indentation (consistently use tabs)
Merge pull request #2046 from rubenk/cleanup-dpdk-detection
Cleanup dpdk detection
Cleanup dpdk detection
Merge remote-tracking branch 'github/pr/2040'
Fix Lua header detection
checking lua.h usability... yes
checking lua.h presence... no
configure: WARNING: lua.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: lua.h: proceeding with the compiler's result
checking for lua.h... yes
checking lauxlib.h usability... yes
checking lauxlib.h presence... no
configure: WARNING: lauxlib.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: lauxlib.h: proceeding with the compiler's result
checking for lauxlib.h... yes
checking lualib.h usability... yes
checking lualib.h presence... no
configure: WARNING: lualib.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: lualib.h: proceeding with the compiler's result
checking for lualib.h... yes
AC_CHECK_HEADERS uses the preprocessor directly, which doesn't use
CFLAGS. Fixes detection of the lua-5.1 compat libs on Fedora Rawhide.
checking lua.h usability... yes
checking lua.h presence... no
configure: WARNING: lua.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: lua.h: proceeding with the compiler's result
checking for lua.h... yes
checking lauxlib.h usability... yes
checking lauxlib.h presence... no
configure: WARNING: lauxlib.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: lauxlib.h: proceeding with the compiler's result
checking for lauxlib.h... yes
checking lualib.h usability... yes
checking lualib.h presence... no
configure: WARNING: lualib.h: accepted by the compiler, rejected by the
preprocessor!
configure: WARNING: lualib.h: proceeding with the compiler's result
checking for lualib.h... yes
AC_CHECK_HEADERS uses the preprocessor directly, which doesn't use
CFLAGS. Fixes detection of the lua-5.1 compat libs on Fedora Rawhide.
src/daemon/plugin.c: Address review comments.
Rewrite the check for libdpdk
This fixes various issues, like us not picking up the right include path
and testing for the library in the wrong location.
We now allow you to override the CPPFLAGS and LDFLAGS by doing
./configure LIBDPDK_CPPFLAGS="-I/path/to/your/libdpdk_headers" LIBDPDK_LDFLAGS="-L/path/to/your/libdpdk.so -Wl,--no-as-needed"
The nice thing is that these settings are now logged and remembered
between configure runs.
This fixes various issues, like us not picking up the right include path
and testing for the library in the wrong location.
We now allow you to override the CPPFLAGS and LDFLAGS by doing
./configure LIBDPDK_CPPFLAGS="-I/path/to/your/libdpdk_headers" LIBDPDK_LDFLAGS="-L/path/to/your/libdpdk.so -Wl,--no-as-needed"
The nice thing is that these settings are now logged and remembered
between configure runs.
Enable detection of libdpdk by default
Otherwise we never exercise this check and it will just bitrot.
Otherwise we never exercise this check and it will just bitrot.
dpdkstat: Added support for dpdk 16.07.
configure.ac: Improve detection and usage of pthread_set_name_np(3).
contrib/README: sort items
contrib/docker: add instructions to README
contrib/docker: add minimal example config file
contrib/docker: add basic Dockerfile & config files
contrib/docker: add LD_PRELOAD wrapper to fopen/open/opendir syscalls
mqtt, gps: add name parameter to plugin_thread_create()
Explain the need for _GNU_SOURCE
Add support to set the thread name.
write_prometheus plugin: Escape label values.
Fixes: #2035
Fixes: #2035
Merge remote-tracking branch 'github/pr/1967'
write_prometheus plugin: Use the "static" macro to specify PROMETHEUS_DEFAULT_STALENESS_DELTA.
Fixes:
write_prometheus.c:56:1: error: initializer element is not constant
static cdtime_t staleness_delta = PROMETHEUS_DEFAULT_STALENESS_DELTA;
^
Fixes:
write_prometheus.c:56:1: error: initializer element is not constant
static cdtime_t staleness_delta = PROMETHEUS_DEFAULT_STALENESS_DELTA;
^
write_prometheus plugin: Don't allocate metric families in prom_missing().
write_prometheus: add support for libmicrohttpd 0.9.45+
Add switch on MHD_VERSION to support both legacy and modern MHD functions.
`MHD_create_response_from_data()` is deprecated since libmicrohttpd
0.9.5 and makes the build fail since 0.9.45.
Add switch on MHD_VERSION to support both legacy and modern MHD functions.
`MHD_create_response_from_data()` is deprecated since libmicrohttpd
0.9.5 and makes the build fail since 0.9.45.
write_prometheus plugin: Improve performance of metric_cmp().
This function is a hotspot because it is used by bsearch() to look up
metrics in a metric family. This simple (though non-obvious) change
brings prom_write() complexity down from 3000 instructions/call to 2640
instructions/call, i.e. a 12% improvement.
This function is a hotspot because it is used by bsearch() to look up
metrics in a metric family. This simple (though non-obvious) change
brings prom_write() complexity down from 3000 instructions/call to 2640
instructions/call, i.e. a 12% improvement.
write_prometheus plugin: Optimize metric_family_get_metric().
Profiling showed that prom_write() spent 73% of its time in this
function. 36% of time was spent in metric_create() and 19% was spent in
metric_destroy().
This patch replaces these two calls by a stack allocation, reducing the
time prom_write() spends in metric_family_get_metric() to 42%.
Profiling showed that prom_write() spent 73% of its time in this
function. 36% of time was spent in metric_create() and 19% was spent in
metric_destroy().
This patch replaces these two calls by a stack allocation, reducing the
time prom_write() spends in metric_family_get_metric() to 42%.
write_prometheus plugin: New plugin for exposing metrics to Prometheus.
collectd-tg: Fix sleep interval.
nanosleep() was called with the current time, meaning that it would
block for years. This correctly subtracts "now" from the metric's time
and only sleeps for the duration between the two.
nanosleep() was called with the current time, meaning that it would
block for years. This correctly subtracts "now" from the metric's time
and only sleeps for the duration between the two.
Merge branch 'collectd-5.6'
Merge branch 'collectd-5.5' into collectd-5.6
Merge remote-tracking branch 'github/pr/2026' into collectd-5.5
apcups plugin: Skip metrics not reported by the UPS.
Fixes: #2025
Fixes: #2025
write_kafka plugin: Fix support for librdkafka 0.9.0.
Use rd_kafka_set_logger() only when rd_kafka_conf_set_log_cb() is not
available:
write_kafka.c: In function 'kafka_handle':
write_kafka.c:119:6: error: 'rd_kafka_set_logger' is deprecated (declared at /usr/local/include/librdkafka/rdkafka.h:2400) [-Werror=deprecated-declarations]
rd_kafka_set_logger(ctx->kafka, kafka_log);
^
Fixes: #2029
Use rd_kafka_set_logger() only when rd_kafka_conf_set_log_cb() is not
available:
write_kafka.c: In function 'kafka_handle':
write_kafka.c:119:6: error: 'rd_kafka_set_logger' is deprecated (declared at /usr/local/include/librdkafka/rdkafka.h:2400) [-Werror=deprecated-declarations]
rd_kafka_set_logger(ctx->kafka, kafka_log);
^
Fixes: #2029
Merge remote-tracking branch 'github/pr/2027' into collectd-5.5
modbus plugin: Avoid sizeof(struct sockaddr).
The previous version failed on FreeBSD with:
modbus.c: In function 'mb_read_data':
modbus.c:476:21: error: storage size of 'sockaddr' isn't known
struct sockaddr sockaddr;
^
The previous version failed on FreeBSD with:
modbus.c: In function 'mb_read_data':
modbus.c:476:21: error: storage size of 'sockaddr' isn't known
struct sockaddr sockaddr;
^
src/write_riemann_threshold.h: Various fixes.
* Add license header. Simply copied src/write_riemann_threadhold.c's license.
* Make header self-contained. "plugin.h" is needed for data_source_t and
value_list_t.
* Add names to the parameters and write documentation for the only exported
function.
* Add license header. Simply copied src/write_riemann_threadhold.c's license.
* Make header self-contained. "plugin.h" is needed for data_source_t and
value_list_t.
* Add names to the parameters and write documentation for the only exported
function.