vitalif
/
etcd
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

*: add debugging metrics

release-3.0
Xiang Li 2016-04-25 15:32:52 -07:00
parent 7161eeed8b
commit 67645095e9
7 changed files with 98 additions and 99 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

131
Documentation/metrics.md
View File

@ -1,46 +1,18 @@
# Metrics
**NOTE: The metrics feature is considered experimental. We may add/change/remove metrics without warning in future releases.**
etcd uses [Prometheus][prometheus] for metrics reporting. The metrics can be used for real-time monitoring and debugging. etcd does not persist its metrics; if a member restarts, the metrics will be reset.
etcd uses [Prometheus][prometheus] for metrics reporting in the server. The metrics can be used for real-time monitoring and debugging.
etcd only stores these data in memory. If a member restarts, metrics will reset.
The simplest way to see the available metrics is to cURL the metrics endpoint `/metrics` of etcd. The format is described [here](http://prometheus.io/docs/instrumenting/exposition_formats/).
The simplest way to see the available metrics is to cURL the metrics endpoint `/metrics`. The format is described [here](http://prometheus.io/docs/instrumenting/exposition_formats/).
Follow the [Prometheus getting started doc][prometheus-getting-started] to spin up a Prometheus server to collect etcd metrics.
The naming of metrics follows the suggested [best practice of Prometheus][prometheus-naming]. A metric name has an `etcd` prefix as its namespace and a subsystem prefix (for example `wal` and `etcdserver`).
The naming of metrics follows the suggested [Prometheus best practices][prometheus-naming]. A metric name has an `etcd` or `etcd_debugging` prefix as its namespace and a subsystem prefix (for example `wal` and `etcdserver`).
etcd now exposes the following metrics:
## etcd namespace metrics
## etcdserver
The metrics under the `etcd` prefix are for monitoring and alerting. They are stable high level metrics. If there is any change of these metrics, it will be included in release notes.
| Name | Description | Type |
|-----------------------------------------|--------------------------------------------------|-----------|
| file_descriptors_used_total | The total number of file descriptors used | Gauge |
| proposal_durations_seconds | The latency distributions of committing proposal | Histogram |
| pending_proposal_total | The total number of pending proposals | Gauge |
| proposal_failed_total | The total number of failed proposals | Counter |
High file descriptors (`file_descriptors_used_total`) usage (near the file descriptors limitation of the process) indicates a potential out of file descriptors issue. That might cause etcd fails to create new WAL files and panics.
[Proposal][glossary-proposal] durations (`proposal_durations_seconds`) provides a histogram about the proposal commit latency. Latency can be introduced into this process by network and disk IO.
Pending proposal (`pending_proposal_total`) gives you an idea about how many proposal are in the queue and waiting for commit. An increasing pending number indicates a high client load or an unstable cluster.
Failed proposals (`proposal_failed_total`) are normally related to two issues: temporary failures related to a leader election or longer duration downtime caused by a loss of quorum in the cluster.
## wal
| Name | Description | Type |
|------------------------------------|--------------------------------------------------|-----------|
| fsync_durations_seconds | The latency distributions of fsync called by wal | Histogram |
| last_index_saved | The index of the last entry saved by wal | Gauge |
Abnormally high fsync duration (`fsync_durations_seconds`) indicates disk issues and might cause the cluster to be unstable.
## http requests
### http requests
These metrics describe the serving of requests (non-watch events) served by etcd members in non-proxy mode: total
incoming requests, request failures and processing latency (inc. raft rounds for storage). They are useful for tracking
@ -71,38 +43,9 @@ Example Prometheus queries that may be useful from these metrics (across all etc
Show the 0.90-tile latency (in seconds) of read/write (respectively) event handling across all members, with a window of `5m`.
## snapshot
### proxy
| Name | Description | Type |
|--------------------------------------------|------------------------------------------------------------|-----------|
| snapshot_save_total_durations_seconds | The total latency distributions of save called by snapshot | Histogram |
Abnormally high snapshot duration (`snapshot_save_total_durations_seconds`) indicates disk issues and might cause the cluster to be unstable.
## rafthttp
| Name | Description | Type | Labels |
|-----------------------------------|--------------------------------------------|--------------|--------------------------------|
| message_sent_latency_seconds | The latency distributions of messages sent | HistogramVec | sendingType, msgType, remoteID |
| message_sent_failed_total | The total number of failed messages sent | Summary | sendingType, msgType, remoteID |
Abnormally high message duration (`message_sent_latency_seconds`) indicates network issues and might cause the cluster to be unstable.
An increase in message failures (`message_sent_failed_total`) indicates more severe network issues and might cause the cluster to be unstable.
Label `sendingType` is the connection type to send messages. `message`, `msgapp` and `msgappv2` use HTTP streaming, while `pipeline` does HTTP request for each message.
Label `msgType` is the type of raft message. `MsgApp` is log replication message; `MsgSnap` is snapshot install message; `MsgProp` is proposal forward message; the others are used to maintain raft internal status. If you have a large snapshot, you would expect a long msgSnap sending latency. For other types of messages, you would expect low latency, which is comparable to your ping latency if you have enough network bandwidth.
Label `remoteID` is the member ID of the message destination.
## proxy
etcd members operating in proxy mode do not do store operations. They forward all requests
to cluster instances.
etcd members operating in proxy mode do not directly perform store operations. They forward all requests to cluster instances.
Tracking the rate of requests coming from a proxy allows one to pin down which machine is performing most reads/writes.
@ -128,7 +71,63 @@ Example Prometheus queries that may be useful from these metrics (across all etc
Number of failed request on the proxy. This should be 0, spikes here indicate connectivity issues to etcd cluster.
## etcd_debugging namespace metrics
The metrics under the `etcd_debugging` prefix are for debugging. They are very implementation dependent and volatile. They might be changed or removed without any warning in new etcd releases. Some of the metrics might be moved to the `etcd` prefix when they become more stable.
### etcdserver
| Name | Description | Type |
|-----------------------------------------|--------------------------------------------------|-----------|
| file_descriptors_used_total | The total number of file descriptors used | Gauge |
| proposal_durations_seconds | The latency distributions of committing proposal | Histogram |
| pending_proposal_total | The total number of pending proposals | Gauge |
| proposal_failed_total | The total number of failed proposals | Counter |
Heavy file descriptor (`file_descriptors_used_total`) usage (i.e., near the process's file descriptor limit) indicates a potential file descriptor exhaustion issue. If the file descriptors are exhausted, etcd may panic because it cannot create new WAL files.
[Proposal][glossary-proposal] durations (`proposal_durations_seconds`) provides a proposal commit latency histogram. The reported latency reflects network and disk IO delays in etcd.
Pending proposal (`pending_proposal_total`) indicates how many proposals are queued for commit. A rising pending proposal total suggests there is a high client load or the cluster is unstable.
Failed proposals (`proposal_failed_total`) are normally related to two issues: temporary failures related to a leader election or longer duration downtime caused by a loss of quorum in the cluster.
### wal
| Name | Description | Type |
|------------------------------------|--------------------------------------------------|-----------|
| fsync_durations_seconds | The latency distributions of fsync called by wal | Histogram |
| last_index_saved | The index of the last entry saved by wal | Gauge |
Abnormally high fsync duration (`fsync_durations_seconds`) indicates disk issues and might cause the cluster to be unstable.
### snapshot
| Name | Description | Type |
|--------------------------------------------|------------------------------------------------------------|-----------|
| snapshot_save_total_durations_seconds | The total latency distributions of save called by snapshot | Histogram |
Abnormally high snapshot duration (`snapshot_save_total_durations_seconds`) indicates disk issues and might cause the cluster to be unstable.
### rafthttp
| Name | Description | Type | Labels |
|-----------------------------------|--------------------------------------------|--------------|--------------------------------|
| message_sent_latency_seconds | The latency distributions of messages sent | HistogramVec | sendingType, msgType, remoteID |
| message_sent_failed_total | The total number of failed messages sent | Summary | sendingType, msgType, remoteID |
Abnormally high message duration (`message_sent_latency_seconds`) indicates network issues and might cause the cluster to be unstable.
An increase in message failures (`message_sent_failed_total`) indicates more severe network issues and might cause the cluster to be unstable.
Label `sendingType` is the connection type to send messages. `message`, `msgapp` and `msgappv2` use HTTP streaming, while `pipeline` does HTTP request for each message.
Label `msgType` is the type of raft message. `MsgApp` is log replication messages; `MsgSnap` is snapshot install messages; `MsgProp` is proposal forward messages; the others maintain internal raft status. Given large snapshots, a lengthy msgSnap transmission latency should be expected. For other types of messages, given enough network bandwidth, latencies comparable to ping latency should be expected.
Label `remoteID` is the member ID of the message destination.
[glossary-proposal]: glossary.md#proposal
[prometheus]: http://prometheus.io/
[prometheus-getting-started](http://prometheus.io/docs/introduction/getting_started/)
[prometheus-getting-started]: http://prometheus.io/docs/introduction/getting_started/
[prometheus-naming]: http://prometheus.io/docs/practices/naming/

8
etcdserver/metrics.go
View File

@ -24,14 +24,14 @@ import (
var (
// TODO: with label in v3?
proposeDurations = prometheus.NewHistogram(prometheus.HistogramOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "server",
Name: "proposal_durations_seconds",
Help: "The latency distributions of committing proposal.",
Buckets: prometheus.ExponentialBuckets(0.001, 2, 14),
})
proposePending = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "server",
Name: "pending_proposal_total",
Help: "The total number of pending proposals.",
@ -39,14 +39,14 @@ var (
// This is number of proposal failed in client's view.
// The proposal might be later got committed in raft.
proposeFailed = prometheus.NewCounter(prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "server",
Name: "proposal_failed_total",
Help: "The total number of failed proposals.",
})
fileDescriptorUsed = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "server",
Name: "file_descriptors_used_total",
Help: "The total number of file descriptors used.",

28
mvcc/metrics.go
View File

@ -21,7 +21,7 @@ import (
var (
rangeCounter = prometheus.NewCounter(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "range_total",
Help: "Total number of ranges seen by this member.",
@ -29,7 +29,7 @@ var (
putCounter = prometheus.NewCounter(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "put_total",
Help: "Total number of puts seen by this member.",
@ -37,7 +37,7 @@ var (
deleteCounter = prometheus.NewCounter(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "delete_total",
Help: "Total number of deletes seen by this member.",
@ -45,7 +45,7 @@ var (
txnCounter = prometheus.NewCounter(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "txn_total",
Help: "Total number of txns seen by this member.",
@ -53,7 +53,7 @@ var (
keysGauge = prometheus.NewGauge(
prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "keys_total",
Help: "Total number of keys.",
@ -61,7 +61,7 @@ var (
watchStreamGauge = prometheus.NewGauge(
prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "watch_stream_total",
Help: "Total number of watch streams.",
@ -69,7 +69,7 @@ var (
watcherGauge = prometheus.NewGauge(
prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "watcher_total",
Help: "Total number of watchers.",
@ -77,7 +77,7 @@ var (
slowWatcherGauge = prometheus.NewGauge(
prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "slow_watcher_total",
Help: "Total number of unsynced slow watchers.",
@ -85,7 +85,7 @@ var (
totalEventsCounter = prometheus.NewCounter(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "events_total",
Help: "Total number of events sent by this member.",
@ -93,7 +93,7 @@ var (
pendingEventsGauge = prometheus.NewGauge(
prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "pending_events_total",
Help: "Total number of pending events to be sent.",
@ -101,7 +101,7 @@ var (
indexCompactionPauseDurations = prometheus.NewHistogram(
prometheus.HistogramOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "index_compaction_pause_duration_milliseconds",
Help: "Bucketed histogram of index compaction pause duration.",
@ -111,7 +111,7 @@ var (
dbCompactionPauseDurations = prometheus.NewHistogram(
prometheus.HistogramOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "db_compaction_pause_duration_milliseconds",
Help: "Bucketed histogram of db compaction pause duration.",
@ -121,7 +121,7 @@ var (
dbCompactionTotalDurations = prometheus.NewHistogram(
prometheus.HistogramOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "db_compaction_total_duration_milliseconds",
Help: "Bucketed histogram of db compaction total duration.",
@ -130,7 +130,7 @@ var (
})
dbTotalSize = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "mvcc",
Name: "db_total_size_in_bytes",
Help: "Total size of the underlying database in bytes.",

4
rafthttp/metrics.go
View File

@ -29,7 +29,7 @@ var (
// time range than other type of messages.
msgSentDuration = prometheus.NewHistogramVec(
prometheus.HistogramOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "rafthttp",
Name: "message_sent_latency_seconds",
Help: "message sent latency distributions.",
@ -39,7 +39,7 @@ var (
)
msgSentFailed = prometheus.NewCounterVec(prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "rafthttp",
Name: "message_sent_failed_total",
Help: "The total number of failed messages sent.",

8
snap/metrics.go
View File

@ -19,16 +19,16 @@ import "github.com/prometheus/client_golang/prometheus"
var (
// TODO: save_fsync latency?
saveDurations = prometheus.NewHistogram(prometheus.HistogramOpts{
Namespace: "etcd",
Subsystem: "snapshot",
Namespace: "etcd_debugging",
Subsystem: "snap",
Name: "save_total_durations_seconds",
Help: "The total latency distributions of save called by snapshot.",
Buckets: prometheus.ExponentialBuckets(0.001, 2, 14),
})
marshallingDurations = prometheus.NewHistogram(prometheus.HistogramOpts{
Namespace: "etcd",
Subsystem: "snapshot",
Namespace: "etcd_debugging",
Subsystem: "snap",
Name: "save_marshalling_durations_seconds",
Help: "The marshalling cost distributions of save called by snapshot.",
Buckets: prometheus.ExponentialBuckets(0.001, 2, 14),

14
store/metrics.go
View File

@ -26,7 +26,7 @@ import (
var (
readCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "store",
Name: "reads_total",
Help: "Total number of reads action by (get/getRecursive), local to this member.",
@ -34,7 +34,7 @@ var (
writeCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "store",
Name: "writes_total",
Help: "Total number of writes (e.g. set/compareAndDelete) seen by this member.",
@ -42,7 +42,7 @@ var (
readFailedCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "store",
Name: "reads_failed_total",
Help: "Failed read actions by (get/getRecursive), local to this member.",
@ -50,7 +50,7 @@ var (
writeFailedCounter = prometheus.NewCounterVec(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "store",
Name: "writes_failed_total",
Help: "Failed write actions (e.g. set/compareAndDelete), seen by this member.",
@ -58,7 +58,7 @@ var (
expireCounter = prometheus.NewCounter(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "store",
Name: "expires_total",
Help: "Total number of expired keys.",
@ -66,7 +66,7 @@ var (
watchRequests = prometheus.NewCounter(
prometheus.CounterOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "store",
Name: "watch_requests_total",
Help: "Total number of incoming watch requests (new or reestablished).",
@ -74,7 +74,7 @@ var (
watcherCount = prometheus.NewGauge(
prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "store",
Name: "watchers",
Help: "Count of currently active watchers.",

4
wal/metrics.go
View File

@ -18,14 +18,14 @@ import "github.com/prometheus/client_golang/prometheus"
var (
syncDurations = prometheus.NewHistogram(prometheus.HistogramOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "wal",
Name: "fsync_durations_seconds",
Help: "The latency distributions of fsync called by wal.",
Buckets: prometheus.ExponentialBuckets(0.001, 2, 14),
})
lastIndexSaved = prometheus.NewGauge(prometheus.GaugeOpts{
Namespace: "etcd",
Namespace: "etcd_debugging",
Subsystem: "wal",
Name: "last_index_saved",
Help: "The index of the last entry saved by wal.",