Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Add systemd notify support to Agent. Resolves: #7028 #9802

Merged
merged 5 commits into from
Aug 29, 2022
Merged

Add systemd notify support to Agent. Resolves: #7028 #9802

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
d0d28b0
Server systemd notify added similar to Consul. Resolves: #7028
aphorise Aug 23, 2020
2da19a1
Server systemd notify added similar to Consul for Vault agent & serve…
aphorise Aug 23, 2020
f5c6866
systemd agent start & shutdown added.
aphorise Sep 14, 2021
234d486
Merge remote-tracking branch 'upstream/main' into server-systemd-notify
aphorise Nov 25, 2021
d7559f9
Merge branch 'main' into server-systemd-notify
aphorise Jul 1, 2022
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions changelog/9802.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,3 @@
```release-note:improvement
agent: Send notifications to systemd on start and stop.
```
20 changes: 20 additions & 0 deletions command/agent.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,6 +16,7 @@ import (
"sync"
"time"

systemd "github.com/coreos/go-systemd/daemon"
log "github.com/hashicorp/go-hclog"
"github.com/hashicorp/go-secure-stdlib/gatedwriter"
"github.com/hashicorp/vault/api"
Expand Down Expand Up @@ -794,13 +795,16 @@ func (c *AgentCommand) Run(args []string) int {
select {
case <-c.ShutdownCh:
c.UI.Output("==> Vault agent shutdown triggered")
// Notify systemd that the server is shutting down
c.notifySystemd(systemd.SdNotifyStopping)
// Let the lease cache know this is a shutdown; no need to evict
// everything
if leaseCache != nil {
leaseCache.SetShuttingDown(true)
}
return nil
case <-ctx.Done():
c.notifySystemd(systemd.SdNotifyStopping)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shouldn't we also do this in the case above, when we were asked to shutdown?

Copy link
Contributor Author

@aphorise aphorise Aug 24, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

hey @ncabatoff I'm not comprehending - does systemd.SdNotifyStopping need to be included in any other case other than these two places?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If any one of the two signals are the only mode of exit (outside Windows) then I think these two later here would suffice and void any repeats - but I'm not entirely sure TBH.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Maybe it's simpler just to add the notifySystemd call to the end of the func, or to a defer.

Copy link
Contributor Author

@aphorise aphorise Aug 24, 2022 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is there any chance that an end or defer function may take longer or not ever reached? - just thinking aloud in case of processor monitors or other cases an agent may be forced quit and never actually ending.

Anyway @ncabatoff any chance you can do a PR if you think a defer may be better.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Defer will always fire when the func returns, and it shouldn't be any slower than doing it just before a return statement.

return nil
case <-winsvc.ShutdownChannel():
return nil
Expand Down Expand Up @@ -928,6 +932,9 @@ func (c *AgentCommand) Run(args []string) int {
return 1
}

// Notify systemd that the server is ready (if applicable)
c.notifySystemd(systemd.SdNotifyReady)
Copy link
Contributor Author

@aphorise aphorise Sep 14, 2021 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I tried adding this a bit later after the run condition which may be more valid to NotifyReady as and when it's actually ready - however since the time it may take for that is not as easily determinable then getting an earlier start notification coupled with any follow-up stop is probably better than just getting NotifyStopping without ever seeing NotifyReady


defer func() {
if err := c.removePidFile(config.PidFile); err != nil {
c.UI.Error(fmt.Sprintf("Error deleting the PID file: %s", err))
Expand Down Expand Up @@ -958,6 +965,19 @@ func verifyRequestHeader(handler http.Handler) http.Handler {
})
}

func (c *AgentCommand) notifySystemd(status string) {
sent, err := systemd.SdNotify(false, status)
if err != nil {
c.logger.Error("error notifying systemd", "error", err)
} else {
if sent {
c.logger.Debug("sent systemd notification", "notification", status)
} else {
c.logger.Debug("would have sent systemd notification (systemd not present)", "notification", status)
}
}
}

func (c *AgentCommand) setStringFlag(f *FlagSets, configVal string, fVar *StringVar) {
var isFlagSet bool
f.Visit(func(f *flag.Flag) {
Expand Down