-
Notifications
You must be signed in to change notification settings - Fork 7
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.
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
feat: proposal pre/post test steps #210
Conversation
Looks good. Left some minor comments. |
README.md
Outdated
@@ -70,6 +70,7 @@ If the proposal is _pending_ and does not yet have a number, use a letter. The p | |||
- `package.json` specifies what kind of proposal it is in a `agoricProposal` field. If it's a "Software Upgrade Proposal" it also includes additional parameters. | |||
- `use.sh` is the script that will be run in the USE stage of the build | |||
- `test.sh` is the script that will be _included_ in the TEST stage of the build, and run in CI | |||
- `prepare-test.sh` is an optional script which can be used to run setup steps _inside_ the container |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
in this PR would you revise the name? Please clarify that it runs within the test image before agd starts. test-before-agd.sh
would be clear.
This readme must also mention the two new hooks.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think the hook name should not have any mention of agd
. It should just be an indication of at which point in the pipeline this hook fits and prepare-test.sh
I think conveys that. We could rename it to pre-test.sh
as well but we then have to distinguish on hook names running outside and inside the container. The current PR uses pre-test.sh
name which run on the host.
What do you suggest?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It should just be an indication of at which point in the pipeline this hook fits
Agree
and prepare-test.sh I think conveys that.
Disagree. prepare.sh
works very differently,
PREPARE For upgrade proposals: submits the proposal, votes on it, runs to halt for the next stage
prepare-test.sh
sounds like it would be a PREPARE step but for tests. However it runs in the same image as test.sh
. It's just like test.sh
but runs before agd starts.
I'll argue firmly that we need a different name than prepare-test.sh
. As you point out pre
and post
are taken. I think those should be more explicit that they don't run within the image: before-test-run.sh
and after-test-run.sh
. ("Before" and "after" more closely match testing hook names like Ava has.)
That would leave pre-test.sh
but that could still be easily confused. I would suggest setup-test.sh
or test-before-chain.sh
.
If we should discuss more let's do it synchronously or in a call.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I agree with before-test-run.sh
and after-test-run.sh
for host and setup-test.sh
(and possibly teardown-test.sh
) for image so we can lock them if you agree as well.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
🤝
I like your "host" term though. I'd suggest putting the host scripts in host
or some other subdir so they're not copied to the image (using --exclude
). That way you can modify them without invalidating it.
const proposalPath = `${root}/proposals/${proposal.path}`; | ||
|
||
if (fileExists(`${proposalPath}/pre_test.sh`)) | ||
execSync(`/bin/bash ${proposalPath}/pre_test.sh`, { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
the scripts have shebang. the caller should not be concerned with how they're interpreted.
execSync(`/bin/bash ${proposalPath}/pre_test.sh`, { | |
execSync(`${proposalPath}/pre_test.sh`, { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
LGTM! Good documentation.
I left one code factoring suggestion at your discretion.
fileExists(afterHookScriptPath) && | ||
execSync(afterHookScriptPath, { stdio: 'inherit' }); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
consider having a helper,
fileExists(afterHookScriptPath) && | |
execSync(afterHookScriptPath, { stdio: 'inherit' }); | |
execHostScriptIfPresent('after-test-run.sh'); |
Then the above three consts don't need to be defined and when someone searched after-test-run
they can see easily where in the sequence it runs.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Please add running the hooks in case we're in debug
mode too.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for making the switch to a tmp file. Let's make sure things are cleaned up, and that we use less manual naming of the tmp file.
import { ProposalInfo, imageNameForProposal } from './proposals.js'; | ||
|
||
const createMessageFile = () => { | ||
const messageFileName = `${new Date().getTime()}.tmp`; | ||
const messageFilePath = `/tmp/${messageFileName}`; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I wasn't sure adding a dependency of tmp
package was really needed here
Added
process.env.MESSAGE_FILE_PATH = messageFilePath; | ||
|
||
executeHostScriptIfPresent(proposal, 'before-test-run.sh'); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it'd be preferable to plumb the messageFilePath
into executeHostScriptIfPresent
and for that spawn to use the modified env of the child process only. Mutating the current process env goes against best practices.
You can either spread the current env, or use inheritance ({__proto__: process.env, MESSAGE_FILE_PATH: messageFilePath}
)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done
}; | ||
|
||
export const debugTestImage = (proposal: ProposalInfo) => { | ||
executeHostScriptIfPresent(proposal, 'before-test-run.sh'); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To clarify, we don't expect tests run in debug mode to have a message file? Is there not a risk that host scripts would expect a message file and break if not available?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Valid point
Done
const { HOME } = env; | ||
|
||
const containerFilePath = `/root/${fileName}`; | ||
const filePath = `${HOME}/${fileName}`; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
My concern was 2 fold:
- Collisions / confusion between multiple test runs. In particular multiple tests of the same
synthetic-chain test
invocation. But to a lesser extent confusion about which file belonged to which invocation, if for some reason the test would not run correctly, but not sufficiently break to report a failure, and the subsequent host hook would then act on unrelated data (if the previous run had not cleaned up correctly) - garbage left behind. The purpose of these files are temporary while the test is running. We should clean them up after completion of the test. And since these are temporary in nature, they belong in the temp folder.
I agree that correctly cleaning up after the test would mitigate both the concerns, but in general I prefer to respect the semantics of the file system.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I like the new generate copy command.
I had missed that a previous commit changed the network bind to use the host interfaces, any reason for that change?
}) => { | ||
const { name: messageFilePath, removeCallback: removeTempFileCallback } = | ||
createMessageFile(proposal); | ||
const messageFileName = basename(messageFilePath); | ||
|
||
const containerFilePath = `/root/${messageFileName}`; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I thought we would have the message file in the container use a static name (not even related to the proposal name) to avoid having to plumb an extra env.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Container file doesn't necessarily need to use the proposal name in it's path, but we still need to pass the host file path to the host start script so that the follower can mount that file in it's container
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Right but the code above will make the container file name dynamic, without any way for the container script to figure out what that name is. Or am I missing something?
`source=${messageFilePath},target=${containerFilePath},type=bind`, | ||
'--network', | ||
'host', | ||
removeContainerOnExit && '--rm', |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
To avoid the .filter(Boolean).map(String)
:
removeContainerOnExit && '--rm', | |
...(removeContainerOnExit ? ['--rm'] : []), |
'--publish', | ||
'1317:1317', | ||
'--publish', | ||
'9090:9090', | ||
'--publish', | ||
'26657:26657', |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why are we losing these port mappings?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Because of the host
network these will still be available
'--network', | ||
'host', |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm pretty sure this is a bad idea. Why do we need to hook onto the host network interfaces?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If we don't use host
network, we would need individual container IPs
Using host
network will make the RPC/Peer available on localhost
(using different ports)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I don't understand. How is different IPs different than different ports?
Also afaik, only the acceptance test needs to be connected to, so couldn't we only target that container?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I never submitted this last week, but for future reference:
We can get the IP address from inside the container using something like
hostname -i
(trimming trailing space). We already need to get the tendermint node-id and probably should get the chain_id (from genesis file). The setup test script can grab those and send over the message file.
{ stdio: 'inherit' }, | ||
); | ||
|
||
executeHostScriptIfPresent( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So at this point our tests are being executed inside container, right?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for restoring the ports config @usmanmani1122 .
Before merging please update the PR description and clean up the commits. I think there are two now:
- feat: proposal pre/post test steps
- fix: exclude node_modules from COPY
(Incidentally, we couldn't exclude node_modules until Docker added the experimental exclude
feature)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks for putting all this in. I second @turadg's request for cleaning up the commits before merging. This PR now does multiple things and I would prefer it not to be squashed, but interactively rebased to tell a clear story (or split into multiple squashed PRs if you prefer).
const { name: messageFilePath, removeCallback: removeTempFileCallback } = | ||
createMessageFile(proposal); | ||
|
||
const containerFilePath = '/root/message-file-path'; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nit:
const containerFilePath = '/root/message-file-path'; | |
const containerFilePath = '/root/host-test-message'; |
'--network', | ||
'host', |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I never submitted this last week, but for future reference:
We can get the IP address from inside the container using something like
hostname -i
(trimming trailing space). We already need to get the tendermint node-id and probably should get the chain_id (from genesis file). The setup test script can grab those and send over the message file.
I think it would be better to create a separate PR for the exclude node_modules fix as there isn't a single commit with just that change |
@usmanmani1122 So what do we plan? Cleaning commits or separate PRs? |
The node_modules change is merged separately. This PR can be squashed and merged now |
@turadg I have merged the |
README.md
Outdated
- `before-test-run.sh` is an optional script which can be used to run setup steps on the _host_ (like starting a follower). It runs _before_ the container launches. It needs to be placed in the **host** folder. | ||
- `after-test-run.sh` is an optional script which can be used to run teardown steps on the _host_ (like stopping a follower). It runs _after_ the container exits. It needs to be placed in the **host** folder. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let's be more explicit about what the host
folder is for. There has been some confusion: Agoric/agoric-sdk#10947 (comment)
- `before-test-run.sh` is an optional script which can be used to run setup steps on the _host_ (like starting a follower). It runs _before_ the container launches. It needs to be placed in the **host** folder. | |
- `after-test-run.sh` is an optional script which can be used to run teardown steps on the _host_ (like stopping a follower). It runs _after_ the container exits. It needs to be placed in the **host** folder. | |
- `host` folder is for ___ | |
- `host/before-test-run.sh` is an optional script which can be used to run setup steps on the _host_ (like starting a follower). It runs _before_ the container launches. | |
- `host/after-test-run.sh` is an optional script which can be used to run teardown steps on the _host_ (like stopping a follower). It runs _after_ the container exits. |
While you're at it a change like this would help,
- `test` folder is for files that should be included only in the `test` image
This is not blocking feedback. If something is delayed by this PR the docs improvements can come later.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Added
closes: #xxxx
@agoric/synthetic-chain library