Add opt-in for index

[gberche-orange]

As a syslog-release release user
In order to filter syslog entries in a similar way than

• bosh ssh instance-group/index
• tail -f /var/vcap/jobs/<job>/*.log
  I need syslog release to issue the instance index in addition to instance id

[ctlong]

Since the current functionality adds the instance ID, you should already be able to bosh ssh using intance-group/guid.

❓ Is there any other reason why the index is desired over the ID?

[gberche-orange]

The uid isn't predictable and tedious to copy/paste, hard to remember.

Using the index is straight forward and hence provides a simpler UX.

[ctlong]

I'm not convinced that this particularly desirable. It seems unnecessarily duplicative and potentially confusing to me. Since there's no guarantee that a particular instance group & index still corresponds to the same VM that emitted the log line it may cause someone to believe they are troubleshooting the correct VM when they are not.

We've been moving away from this concept of an instance index in favor of the instance id for some time. The new, and as yet unused, Cloud Foundry attributes for OTel don't specify it, for instance.

What do you think @chombium?

[chombium]

I agree with @ctlong. When I think of checking something by instance index, I ony use that for checking a config of a job as the configuration for all jobs in a particular instance group is the same. It should not be forgotten that the instance index changes and doesn't point to the same vm when one scales something horizontally or during foundation updates when vms are being replaced.

I would never use instance index to check sole logs from a specific vm, as it is not possible to deterministically target an exact vm.

@gberche-orange Can you tell us a bit more about your use case, apart from ease of use? Could you maybe use something from the existing metadata to search for what you need?