From 9b2dce5763de16e2374428884ec6cf8072256337 Mon Sep 17 00:00:00 2001 From: Ekaterina Lebedeva Date: Thu, 16 Nov 2023 02:48:05 +0300 Subject: [PATCH] [#552] Add status notification to systemd Signed-off-by: Ekaterina Lebedeva --- cmd/frostfs-node/control.go | 17 +++++++++++++++++ debian/frostfs-ir.service | 3 ++- debian/frostfs-storage.service | 3 ++- go.mod | 1 + go.sum | 12 ++++++++++++ internal/logs/logs.go | 1 + 6 files changed, 35 insertions(+), 2 deletions(-) diff --git a/cmd/frostfs-node/control.go b/cmd/frostfs-node/control.go index 4f5c5d780..fcd84db3c 100644 --- a/cmd/frostfs-node/control.go +++ b/cmd/frostfs-node/control.go @@ -2,6 +2,7 @@ package main import ( "context" + "fmt" "net" controlconfig "git.frostfs.info/TrueCloudLab/frostfs-node/cmd/frostfs-node/config/control" @@ -10,6 +11,7 @@ import ( controlSvc "git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/control/server" "git.frostfs.info/TrueCloudLab/frostfs-node/pkg/services/tree" cid "git.frostfs.info/TrueCloudLab/frostfs-sdk-go/container/id" + sysdnotify "github.com/iguanesolutions/go-systemd/v5/notify" "go.uber.org/zap" "google.golang.org/grpc" ) @@ -83,12 +85,14 @@ func (c *cfg) NetmapStatus() control.NetmapStatus { } func (c *cfg) setHealthStatus(st control.HealthStatus) { + c.notifySystemd(st) c.healthStatus.Store(int32(st)) c.metricsCollector.State().SetHealth(int32(st)) } func (c *cfg) compareAndSwapHealthStatus(oldSt, newSt control.HealthStatus) (swapped bool) { if swapped = c.healthStatus.CompareAndSwap(int32(oldSt), int32(newSt)); swapped { + c.notifySystemd(newSt) c.metricsCollector.State().SetHealth(int32(newSt)) } return @@ -96,6 +100,7 @@ func (c *cfg) compareAndSwapHealthStatus(oldSt, newSt control.HealthStatus) (swa func (c *cfg) swapHealthStatus(st control.HealthStatus) (old control.HealthStatus) { old = control.HealthStatus(c.healthStatus.Swap(int32(st))) + c.notifySystemd(st) c.metricsCollector.State().SetHealth(int32(st)) return } @@ -103,3 +108,15 @@ func (c *cfg) swapHealthStatus(st control.HealthStatus) (old control.HealthStatu func (c *cfg) HealthStatus() control.HealthStatus { return control.HealthStatus(c.healthStatus.Load()) } + +func (c *cfg) notifySystemd(st control.HealthStatus) string { + status := fmt.Sprintf("%v, %v", st.Number(), st) + err := sysdnotify.Status(status) + if err == nil { + c.log.Info(fmt.Sprintf("reported STATUS=\"%v\" to systemd", status)) + } + if err != nil { + c.log.Error(logs.FailedToReportStatusToSystemd, zap.String("error", err.Error())) + } + return fmt.Sprintf("%v, %v", st, control.HealthStatus_name[int32(st)]) +} diff --git a/debian/frostfs-ir.service b/debian/frostfs-ir.service index b59295ac5..304017f68 100644 --- a/debian/frostfs-ir.service +++ b/debian/frostfs-ir.service @@ -3,7 +3,8 @@ Description=FrostFS InnerRing node Requires=network.target [Service] -Type=simple +Type=notify +NotifyAccess=all ExecStart=/usr/bin/frostfs-ir --config /etc/frostfs/ir/config.yml User=frostfs-ir Group=frostfs-ir diff --git a/debian/frostfs-storage.service b/debian/frostfs-storage.service index 6e7d1dd5d..573961756 100644 --- a/debian/frostfs-storage.service +++ b/debian/frostfs-storage.service @@ -3,7 +3,8 @@ Description=FrostFS Storage node Requires=network.target [Service] -Type=simple +Type=notify +NotifyAccess=all ExecStart=/usr/bin/frostfs-node --config /etc/frostfs/storage/config.yml User=frostfs-storage Group=frostfs-storage diff --git a/go.mod b/go.mod index 3e79f7ff7..3037a6167 100644 --- a/go.mod +++ b/go.mod @@ -16,6 +16,7 @@ require ( github.com/flynn-archive/go-shlex v0.0.0-20150515145356-3f9db97f8568 github.com/google/uuid v1.4.0 github.com/hashicorp/golang-lru/v2 v2.0.7 + github.com/iguanesolutions/go-systemd/v5 v5.1.1 github.com/klauspost/compress v1.17.4 github.com/mitchellh/go-homedir v1.1.0 github.com/mr-tron/base58 v1.2.0 diff --git a/go.sum b/go.sum index 15ae22422..780bdbcdb 100644 --- a/go.sum +++ b/go.sum @@ -146,6 +146,18 @@ github.com/holiman/uint256 v1.2.4 h1:jUc4Nk8fm9jZabQuqr2JzednajVmBpC+oiTiXZJEApU github.com/holiman/uint256 v1.2.4/go.mod h1:EOMSn4q6Nyt9P6efbI3bueV4e1b3dGlUCXeiRV4ng7E= github.com/hpcloud/tail v1.0.0/go.mod h1:ab1qPbhIpdTxEkNHXyeSf5vhxWSCs/tWer42PpOxQnU= github.com/ianlancetaylor/demangle v0.0.0-20200824232613-28f6c0f3b639/go.mod h1:aSSvb/t6k1mPoxDqO4vJh6VOCGPwU4O0C2/Eqndh1Sc= +github.com/ianlancetaylor/demangle v0.0.0-20230524184225-eabc099b10ab/go.mod h1:gx7rwoVhcfuVKG5uya9Hs3Sxj7EIvldVofAWIUtGouw= +github.com/iguanesolutions/go-systemd/v5 v5.1.1 h1:Hs0Z16knPGCBFnKECrICPh+RQ89Sgy0xyzcalrHMKdw= +github.com/iguanesolutions/go-systemd/v5 v5.1.1/go.mod h1:Quv57scs6S7T0rC6qyLfW20KU/P4p9hrbLPF+ILYrXY= +github.com/imdario/mergo v0.3.5/go.mod h1:2EnlNZ0deacrJVfApfmtdGgDfMuh/nq6Ok1EcJh5FfA= +github.com/imdario/mergo v0.3.6/go.mod h1:2EnlNZ0deacrJVfApfmtdGgDfMuh/nq6Ok1EcJh5FfA= +github.com/imdario/mergo v0.3.8/go.mod h1:2EnlNZ0deacrJVfApfmtdGgDfMuh/nq6Ok1EcJh5FfA= +github.com/imdario/mergo v0.3.10/go.mod h1:jmQim1M+e3UYxmgPu/WyfjB3N3VflVyUjjjwH0dnCYA= +github.com/imdario/mergo v0.3.11/go.mod h1:jmQim1M+e3UYxmgPu/WyfjB3N3VflVyUjjjwH0dnCYA= +github.com/imdario/mergo v0.3.12/go.mod h1:jmQim1M+e3UYxmgPu/WyfjB3N3VflVyUjjjwH0dnCYA= +github.com/imdario/mergo v0.3.13/go.mod h1:4lJ1jqUDcsbIECGy0RUJAXNIhg+6ocWgb1ALK2O4oXg= +github.com/inconshreveable/mousetrap v1.0.0/go.mod h1:PxqpIevigyE2G7u3NXJIT2ANytuPF1OarO4DADm73n8= +github.com/inconshreveable/mousetrap v1.0.1/go.mod h1:vpF70FUmC8bwa3OWnCshd2FqLfsEA9PFc4w1p2J65bw= github.com/inconshreveable/mousetrap v1.1.0 h1:wN+x4NVGpMsO7ErUn/mUI3vEoE6Jt13X2s0bqwp9tc8= github.com/inconshreveable/mousetrap v1.1.0/go.mod h1:vpF70FUmC8bwa3OWnCshd2FqLfsEA9PFc4w1p2J65bw= github.com/ipfs/go-cid v0.4.1 h1:A/T3qGvxi4kpKWWcPC/PgbvDA2bjVLO7n4UeVwnbs/s= diff --git a/internal/logs/logs.go b/internal/logs/logs.go index c996e830a..1775b18cd 100644 --- a/internal/logs/logs.go +++ b/internal/logs/logs.go @@ -554,4 +554,5 @@ const ( BlobovniczaSavingCountersToMetaSuccess = "saving counters to blobovnicza's meta completed successfully" BlobovniczaSavingCountersToMetaFailed = "saving counters to blobovnicza's meta failed" ObjectRemovalFailureExistsInWritecache = "can't remove object: object must be flushed from writecache" + FailedToReportStatusToSystemd = "failed to report status to systemd" )