Prevent service being started/enabled on install for rpms

[sproberts92]

This brings behaviour inline with usual expectations for rpm packages which usually leave this up to the system administrator.

Create a separate postinstall-rpm.sh specifically for rpms and set this as an override for the postinstall script for rpm packages.

This resolves #1334.

[sproberts92]

I am uncertain whether overrides will deep-merge the scripts block. Until I manage to get goreleaser to work locally I'm hoping a CI run can produce a package that I can inspect.

[sproberts92]

Hmm this won't restart on an upgrade either. I will fix that tomorrow night. The original form of the script but using systemd try-restart instead should do the trick.

[atoulme]

this changelog doesn't reflect the changes. You seem to be checking for systemd presence in the system.
Please clarify this changelog.

[sproberts92]

I believe this is an accurate description of the change in behaviour for end users - please also refer to #1334 for more expansion on the motivation here.

All of the logic in the scripts that I am adding already exists in the base postinstall.sh scripts - my change merely creates specific versions with slightly different outcomes for rpms.

For example, for the plain otelcol distribution, the diff between the existing script and my rpm specific override:

--- a/distributions/otelcol/postinstall.sh
+++ b/distributions/otelcol/postinstall-rpm.sh
@@ -7,10 +7,9 @@ if command -v systemctl >/dev/null 2>&1; then
     if [ -d /run/systemd/system ]; then
         systemctl daemon-reload
     fi
-    systemctl enable otelcol.service
     if [ -f /etc/otelcol/config.yaml ]; then
         if [ -d /run/systemd/system ]; then
-            systemctl restart otelcol.service
+            systemctl try-restart otelcol.service
         fi
     fi
 fi

If you think there's a better approach then I'm happy to adjust.

[atoulme]

OK I see. There is boilerplate on that script you can strip away imo ; remove the check for the config file presence. Is checking for /run/systemd/system presence even necessary?

[sproberts92]

I am uncertain whether overrides will deep-merge the scripts block. Until I manage to get goreleaser to work locally I'm hoping a CI run can produce a package that I can inspect.

Figured out how to build the rpms locally now - so yes the config I've used here does deep merge the scripts block:

$ rpm -qlp --scripts otelcol_0.144.0-next_linux_amd64.rpm
preinstall scriptlet (using /bin/sh):
#!/bin/sh

# Copyright The OpenTelemetry Authors
# SPDX-License-Identifier: Apache-2.0

getent passwd otel >/dev/null || useradd --system --user-group --no-create-home --shell /sbin/nologin otel

postinstall scriptlet (using /bin/sh):
#!/bin/sh

# Copyright The OpenTelemetry Authors
# SPDX-License-Identifier: Apache-2.0

if command -v systemctl >/dev/null 2>&1; then
    if [ -d /run/systemd/system ]; then
        systemctl daemon-reload
    fi
    if [ -f /etc/otelcol/config.yaml ]; then
        if [ -d /run/systemd/system ]; then
            systemctl try-restart otelcol.service
        fi
    fi
fi

preuninstall scriptlet (using /bin/sh):
#!/bin/sh

# Copyright The OpenTelemetry Authors
# SPDX-License-Identifier: Apache-2.0

if [ "$1" != "1" ]; then
    if command -v systemctl >/dev/null 2>&1; then
        systemctl stop otelcol.service
        systemctl disable otelcol.service
    fi
fi

/etc/otelcol/config.yaml
/etc/otelcol/otelcol.conf
/lib/systemd/system/otelcol.service
/usr/bin/otelcol

[sproberts92]

The current CI failure looks unrelated to my change, and probably transient. @mowies, any chance you could please re-trigger?

[atoulme]

I reran CI

[sproberts92]

Looks better now, thanks 👍

[atoulme]

why are you checking for this file presence?