Skip to content

Commit 3ca8667

Browse files
authored
Add major upgrade prechecks (zalando#2772)
Don't fail major upgrade (don't set annotation) if replica(s) are not (yet) streaming or replication lag is too high
1 parent c25dc57 commit 3ca8667

File tree

3 files changed

+41
-3
lines changed

3 files changed

+41
-3
lines changed

go.mod

Lines changed: 1 addition & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -22,6 +22,7 @@ require (
2222
)
2323

2424
require (
25+
github.com/Masterminds/semver v1.5.0
2526
github.com/davecgh/go-spew v1.1.1 // indirect
2627
github.com/emicklei/go-restful/v3 v3.11.0 // indirect
2728
github.com/evanphx/json-patch v4.12.0+incompatible // indirect

go.sum

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,3 +1,5 @@
1+
github.com/Masterminds/semver v1.5.0 h1:H65muMkzWKEuNDnfl9d70GUjFniHKHRbFPGBuZ3QEww=
2+
github.com/Masterminds/semver v1.5.0/go.mod h1:MB6lktGJrhw8PrUyiEoblNEGEQ+RzHPF078ddwwvV3Y=
13
github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5 h1:0CwZNZbxp69SHPdPJAN/hZIm0C4OItdklCFmMRWYpio=
24
github.com/armon/go-socks5 v0.0.0-20160902184237-e75332964ef5/go.mod h1:wHh0iHkYZB8zMSxRWpUBQtwG5a7fFgvEO+odwuTv2gs=
35
github.com/aws/aws-sdk-go v1.53.8 h1:eoqGb1WOHIrCFKo1d51cMcnt1ralfLFaEqRkC5Zzv8k=

pkg/cluster/majorversionupgrade.go

Lines changed: 38 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -6,6 +6,7 @@ import (
66
"fmt"
77
"strings"
88

9+
"github.com/Masterminds/semver"
910
"github.com/zalando/postgres-operator/pkg/spec"
1011
"github.com/zalando/postgres-operator/pkg/util"
1112
v1 "k8s.io/api/core/v1"
@@ -170,6 +171,38 @@ func (c *Cluster) majorVersionUpgrade() error {
170171
return nil
171172
}
172173

174+
members, err := c.patroni.GetClusterMembers(masterPod)
175+
if err != nil {
176+
c.logger.Error("could not get cluster members data from Patroni API, skipping major version upgrade")
177+
return err
178+
}
179+
patroniData, err := c.patroni.GetMemberData(masterPod)
180+
if err != nil {
181+
c.logger.Error("could not get members data from Patroni API, skipping major version upgrade")
182+
return err
183+
}
184+
patroniVer, err := semver.NewVersion(patroniData.Patroni.Version)
185+
if err != nil {
186+
c.logger.Error("error parsing Patroni version")
187+
patroniVer, _ = semver.NewVersion("3.0.4")
188+
}
189+
verConstraint, _ := semver.NewConstraint(">= 3.0.4")
190+
checkStreaming, _ := verConstraint.Validate(patroniVer)
191+
192+
for _, member := range members {
193+
if PostgresRole(member.Role) == Leader {
194+
continue
195+
}
196+
if checkStreaming && member.State != "streaming" {
197+
c.logger.Infof("skipping major version upgrade, replica %s is not streaming from primary", member.Name)
198+
return nil
199+
}
200+
if member.Lag > 16*1024*1024 {
201+
c.logger.Infof("skipping major version upgrade, replication lag on member %s is too high", member.Name)
202+
return nil
203+
}
204+
}
205+
173206
isUpgradeSuccess := true
174207
numberOfPods := len(pods)
175208
if allRunning && masterPod != nil {
@@ -187,19 +220,21 @@ func (c *Cluster) majorVersionUpgrade() error {
187220
}
188221

189222
resultIdCheck = strings.TrimSuffix(resultIdCheck, "\n")
190-
var result string
223+
var result, scriptErrMsg string
191224
if resultIdCheck != "0" {
192225
c.logger.Infof("user id was identified as: %s, hence default user is non-root already", resultIdCheck)
193226
result, err = c.ExecCommand(podName, "/bin/bash", "-c", upgradeCommand)
227+
scriptErrMsg, _ = c.ExecCommand(podName, "/bin/bash", "-c", "tail -n 1 last_upgrade.log")
194228
} else {
195229
c.logger.Infof("user id was identified as: %s, using su to reach the postgres user", resultIdCheck)
196230
result, err = c.ExecCommand(podName, "/bin/su", "postgres", "-c", upgradeCommand)
231+
scriptErrMsg, _ = c.ExecCommand(podName, "/bin/bash", "-c", "tail -n 1 last_upgrade.log")
197232
}
198233
if err != nil {
199234
isUpgradeSuccess = false
200235
c.annotatePostgresResource(isUpgradeSuccess)
201-
c.eventRecorder.Eventf(c.GetReference(), v1.EventTypeWarning, "Major Version Upgrade", "upgrade from %d to %d FAILED: %v", c.currentMajorVersion, desiredVersion, err)
202-
return err
236+
c.eventRecorder.Eventf(c.GetReference(), v1.EventTypeWarning, "Major Version Upgrade", "upgrade from %d to %d FAILED: %v", c.currentMajorVersion, desiredVersion, scriptErrMsg)
237+
return fmt.Errorf(scriptErrMsg)
203238
}
204239

205240
c.annotatePostgresResource(isUpgradeSuccess)

0 commit comments

Comments
 (0)