`stratis pool start --unlock-method clevis --name pool1` does NOT find pool

[aanno]

I encounter the following problem with stratis 3.8.3 on fedora 43.

# clevis is configured and hence the following 2 commands succeeds:
stratis pool start --unlock-method clevis --name pool1
stratis pool start --unlock-method clevis --name pool2

# hence the following also works as expected
stratis pool
Name                  Total / Used / Free        Properties                                   UUID   Alerts
pool1    1.91 TiB / 619.75 GiB / 1.30 TiB    Le, Ca, Cr, Op   2304641c-8acc-4bbd-910a-b70d3138d00f         
pool2   878.89 GiB / 970 MiB / 877.94 GiB    Le,~Ca, Cr, Op   99aa74b7-4ff9-411f-b8a4-d49e3397b0b3   

# however I think this is VERY strange and feels like a bug to me
stratis pool start --unlock-method clevis --name pool1
Execution failed:
It appears that you issued an unintended command: Command 'start' assumes that resource 'pool with name pool1' exists but it was not found.
stratis pool start --unlock-method clevis --name pool2
Execution failed:
It appears that you issued an unintended command: Command 'start' assumes that resource 'pool with name pool2' exists but it was not found.

# same problem with uuid
stratis pool start --unlock-method clevis --uuid 2304641c-8acc-4bbd-910a-b70d3138d00f
Execution failed:
It appears that you issued an unintended command: Command 'start' assumes that resource 'pool with UUID 2304641c-8acc-4bbd-910a-b70d3138d00f' exists but it was not found.
stratis pool start --unlock-method clevis --uuid 99aa74b7-4ff9-411f-b8a4-d49e3397b0b3 
Execution failed:
It appears that you issued an unintended command: Command 'start' assumes that resource 'pool with UUID 99aa74b7-4ff9-411f-b8a4-d49e3397b0b3' exists but it was not found.

Version:

stratis daemon version
3.8.6

[coderabbitai]

📝 CodeRabbit Plan Mode

Generate an implementation plan and prompts that you can use with your favorite coding agent.

• Create Plan

Examples

• Example 1
• Example 2

---

🔗 Similar Issues

Related Issues

• Number of pool subcommands is getting too long; we need to split them out before we add additional ones for reencrypt, encrypt, decrypt #1136

👤 Suggested Assignees

• mulkieran

---

🧪 Issue enrichment is currently in open beta.

You can configure auto-planning by selecting labels in the issue_enrichment configuration.

To disable automatic issue enrichment, add the following to your .coderabbit.yaml:

issue_enrichment:
  auto_enrich:
    enabled: false

💬 Have feedback or questions? Drop into our discord or schedule a call!

[mulkieran]

@aanno Assuming the events you describe occur in sequence I think this is merely a problem of error messages that are perhaps too generic to be really helpful to the user.

Once the pool has been started, the CLI attempts to give feedback to the user that the pool has already been started so it is pointless to try to start it again. For a human, this is valuable information, as the intention may have been to start a wholly different pool and the human fumble-fingered.

stratisd tries to be idempotent, so it would just return success. Early on, stratis-cli was also idempotent, but many users complained, so we have it check certain things and return an error if it looks like the user made a mistake.

If you want to give a suggestion for how the error message can be more helpful we would be happy to consider it.

[aanno]

Well, I think having the error message 'Execution failed:
It appears that you issued an unintended command: Command 'start' assumes that resource 'pool with name pool1' exists but it was not found.' on the second use of stratis pool start --unlock-method clevis --name pool1 is completely misleading.

Expected behaviour:

1. Warning instead of error (e.g. the return code should be 0)
2. A message in the lines of: "The resource 'pool with name pool1' has already been started. Hence there is no need to start it again."

[mulkieran]

Well, I think having the error message 'Execution failed: It appears that you issued an unintended command: Command 'start' assumes that resource 'pool with name pool1' exists but it was not found.' on the second use of stratis pool start --unlock-method clevis --name pool1 is completely misleading.

Expected behaviour:

1. Warning instead of error (e.g. the return code should be 0)

2. A message in the lines of: "The resource 'pool with name pool1' has already been started. Hence there is no need to start it again."

We couldn't change the exit code to 0, because we have decided the CLI will not be idempotent. Going with something more specific to the actual situation is likely possible though.

[mulkieran]

Likely best method is changing string information, like "start", to an enum, and then analyzing the situation further when the exception is caught in the reporting module.

[jbaublitz]

@mulkieran I'm slightly confused here because doing the zbus migration, I've been getting a different error message:

It appears that you issued an unintended command: The 'start' action has no effect for resource pool with name testpool

Is this a change to the error message in a more recent version of stratis-cli or is something else going on here?

[mulkieran]

@mulkieran I'm slightly confused here because doing the zbus migration, I've been getting a different error message:

It appears that you issued an unintended command: The 'start' action has no effect for resource pool with name testpool

Is this a change to the error message in a more recent version of stratis-cli or is something else going on here?

@jbaublitz These different messages correspond to different exceptions raised by stratis-cli. Your message corresponds to a StratisCliNoChangeError while the message in the original post corresponds to StratisCliResourceNotFoundError. The message in the original post is the correct one. What that means is that the stopped pool is still being found in the StoppedPools property by stratis-cli in your tests.

[jbaublitz]

I have confirmed that StoppedPools is completely empty both in the report interface and as a property so I do not believe that that is the cause.

[mulkieran]

I have confirmed that StoppedPools is completely empty both in the report interface and as a property so I do not believe that that is the cause.

We realized that a different command-line was causing the code to skip the pre-check in the second case.