Vercel: Alias immediately after a deploy now fails

[SteveALee]

my npm deploy script is

now --public deploy build/ && now alias alwaysinmind

but sometime the alias errors wit ha message that the deployment is not ready.

Would it be possible to make the deploy wait?

If not do you have any other suggestions - eg a portable delay cmd :)

[SteveALee]

In fact sometime the deploy is failing.

$ now --public deploy build/
> Deploying C:\projects\aim-web-app\client\build under stevelee
> Using Node.js 8.9.4 (default)
> Ready! https://alwaysinmind-qntfvyiiaa.now.sh [2s]
> Initializing…

and no more status outputs to console

[steida]

The same here. From yesterday I am not able to deploy.

[steida]

[steida]

I thought it was now error, but it was mine error. https://zeit.co/steida/este/juubjtgoxn/logs
The problem of cli is that is will not tell you that deployment has failed.

[SteveALee]

@steida - ah thanks - will check that. The Alais command does give a nice red error.

In fact sometime the deploy is failing.

actually it's not failing it just that static deployments like my SPA frontend don't give all the details that node ones do. I mistakenly took this as a failure.

However the original issue remains. While using alias without specifying the actuall instance to link to is not document as afar as I can tell. It always used to work. Perhaps something now has gone asynchronous so the CLI returns from the previous deploy before it is complete. Or perhasps it just a timing thing and something is slow?

[davidyaha]

The CLI exits successfully after initializing - This started happening to me after updating to 0.9.3

[SteveALee]

Yes, it only recently started happening. I have the WIndows desktop client set to auto update but it's not clear if that also updates the global CLI.

I also installed project locally via yarn as got a red 'new version' warning form the cli. That has a few problems - report in another bug

[SteveALee]

A hacky package.json script to give a portable delay till this command becomes synchronous again

    "pause": "node -e \"((time) => new Promise((resolve) => setTimeout(resolve, time)))(process.argv[1]).then(() => process.exit(0))\" 3000"

[OlliV]

We are investigating the issue. We recently started to enforce that an alias can't be pointed to a deployment that can't handle traffic. Unfortunately it seems these changes had some side effects.

[slorber]

I get this issue too when running the 2 commands one after the other. Previously I was using the deprecated --alias option but it seems removed now

[codemzy]

Also just started getting this problem running now && now alias which worked fine up until a few days ago. It seems like before this problem, I could see the deployment building, installing dependencies etc. Now in the CLI I see Initializing… and then the alias attempt right away.

Upgrading to the latest CLI version same issue. Good you see you guys are on it.

In the meantime, I find if I run the same command 30 seconds later, it sees there are no changes to the code, finds the previous deployment that was not ready before and assigns the alias to it.

[SteveALee]

@codemzy I was using 3 Secs but that still fails a small amount of times.

Once I use proper continuous deployment with CI I won't mind much but while it's manual, an extra 30 secs wait is a pain

[chrillo]

as a workaround this works for me, it sure is not pretty though

now  && sleep 60 && now alias && now rm project --safe --yes

[spences10]

Does it work in a Travis pipeline @chrillo ? I tried @SteveALee's pause script for 30 seconds and it still skipped the alias

[chrillo]

@spences10 I can't say for sure but the sleep command is available on most unix systems so I'd imagine it is

[SteveALee]

Sleep got removed from windows and replaced with timeout ages ago.

I find in git for windows bash that timeout always runs the bash version of timeout which is no use. Hence the one line portable node script I posted above.

[SteveALee]

@spences10 try increasing the time constant to say 30000 (30 secs) if you can bear the delay :) I find it occasionally fails with 3 Secs so have increased it in my script

[chrillo]

I found 30 seconds being not enough some times, with 60 I have not experienced any failures.

[spences10]

I'll try with a minute or two, thanks 👍

[SteveALee]

Any advance on 2 mins? :)

[matheuss]

I was investigating this and failed to reproduce it – tried around 15 times with no luck.

Could you guys run now ls | head -6 immediately after the failure to alias, and share the full logs (the output of now && now alias and the output of now ls | head -6)?

[matheuss]

I also added some extra logging on the relevant APIs that will help shed light into what's causing this, specially into @OlliV's suspiciousness.

[SteveALee]

@matheuss sure, will try ASAP :)

[SteveALee]

@matheuss

C:\projects\aim-web-app>yarn deploy:api
yarn run v1.3.2
$ now rm -y alwaysinmindapi
> Success! [2s]
> Deployment alwaysinmindapi-tummapzdqw.now.sh removed
$ now --public deploy api/src && yarn pause && now alias alwaysinmindapi
UPDATE AVAILABLE The latest version of Now CLI is 9.3.3
> Deploying C:\projects\aim-web-app\api\src under stevelee
> Using Node.js 8.9.4 (default)
> Ready! https://alwaysinmindapi-clcopkyejw.now.sh [2s]
> Initializing…
$ rem node -e "((time) => new Promise((resolve) => setTimeout(resolve, time)))(process.argv[1]).then(() => process.exit(0))" 30000
> Assigning alias alwaysinmindapi to deployment...
> Error! Unknown error: Error: The deployment is not ready yet
Error: The deployment is not ready yet
    at module.exports.Now.createAlias.alias.retry (C:\snapshot\now-cli\dist\now.js:3978:21)
    at <anonymous>
    at process._tickCallback (internal/process/next_tick.js:188:7)
Done in 17.98s.

C:\projects\aim-web-app>bash

steve@OD-Surface MINGW64 /c/projects/aim-web-app (master)
$ now ls | head -6
> 2 deployments found under stevelee [980ms]

alwaysinmindapi (1 of 1 total)
 url                                       inst #    state                 age
 alwaysinmindapi-clcopkyejw.now.sh              1    READY                 49s

oh is the new API logging on the client side? I'll update and re run just in case

[SteveALee]

The output with now 9.3.3 is identical even with yarn run deploy:client && now ls | head -6 as a single command

[fridays]

@matheuss I think it's the time before an app listens to a public port when aliasing is not possible

[SteveALee]

If you can;t get it to happen then my guess is latency somehow is causing it - it's unfortunately significant here in UK >700ms. Though you's think that would make things better :)

[kochie]

@SteveALee you can make a now.json file and put the aliases in that. When you run now it should automatically add the aliases to your deployment.

{
    "alias" : [
        "alias.now.sh",
        "second.alias.now.sh"
    ]
}

Not sure if it will fix the reason for your error but it might help you troubleshoot, and it's a lot nicer than writing scripts. :wink:

edit: fixed typo

[SteveALee]

@kochie Oh, OK, will try. I already have a now.json to set the name different to the npm package name. Though 'files' seems misnamed to me! :)

[matheuss]

@kochie files is used to determine which files now should upload – not to determine which aliases to set. alias is for that 🙂

@SteveALee thanks for re-running it. The logs in the server-side were helpful – I'll share more later today.

@fridays aliasing is only possible after the deployment is completed, which only happens after an exposed port is detected.

[SteveALee]

@kochie that does NOT work the, alias is not created

[SteveALee]

@matheuss yes it seemed a strange use of 'files' :)

Glad you got some useful intell!

[kochie]

@SteveALee crap that's my bad! alias is the correct name, my apologies. Thanks for the pickup @matheuss .

[SteveALee]

@kochie heh, NP! But alias setting is only a way of providing the alias name for the alias command when not on the command line so I doubt it will have any impact.

[matheuss]

Should be fixed by now! It was a tricky (but simple) issue that was only affecting requests being served by bru.now.sh. I'm in Brazil, so I'm being served by sfo.now.sh – that's why I had a hard time reproducing it.

@SteveALee could you confirm it doesn't happen anymore? 🙂

[spences10]

Ok, I had now ls | head -6 set on my .travis.yml for this, do I not need to do that now @matheuss? I'm UK based too

[matheuss]

@spences10 shouldn't be needed anymore since I've already identified the issue and fixed it 🙏 Could you also confirm it's not happening anymore? 🙌

[spences10]

@matheuss I'll try now 👍

And thanks! 🙏

[spences10]

🤔

Still the same behaviour for me @matheuss, you want me to add the now ls | head -6 and try again?

[SteveALee]

Wahoo!!!!! It works. I now see those nice little status lines again. I had convinced myself I had imagined them and they only happened for node deployments, not static :)

$ yarn deploy:api && now ls && now alias ls
yarn run v1.3.2
$ now rm -y alwaysinmindapi
> Success! [1s]
> Deployment alwaysinmindapi-pzhrsdteex.now.sh removed
$ now --public deploy api/src && yarn pause && now alias alwaysinmindapi
> Deploying C:\projects\aim-web-app\api\src under stevelee
> Using Node.js 8.9.4 (default)
> Ready! https://alwaysinmindapi-kxmoebtgsz.now.sh [2s]
> Initializing…
> Building
> ▲ npm install
> ⧗ Installing 4 main dependencies…
> ✓ Installed 88 modules [1s]
> ▲ npm start
> > [email protected] start /home/nowuser/src
> > micro
> micro: Accepting connections on port 3000
> Deployment complete!
$ rem node -e "((time) => new Promise((resolve) => setTimeout(resolve, time)))(process.argv[1]).then(() => process.exit(0))" 30000
> Assigning alias alwaysinmindapi to deployment...
> Success! alwaysinmindapi.now.sh now points to alwaysinmindapi-kxmoebtgsz.now.sh! [4s]
Done in 58.42s.
> 2 deployments found under stevelee [922ms]

alwaysinmindapi (1 of 1 total)
 url                                       inst #    state                 age
 alwaysinmindapi-kxmoebtgsz.now.sh              1    READY                 56s

alwaysinmind (1 of 1 total)
 url                                       inst #    state                 age
 alwaysinmind-ovrebfxffj.now.sh                 0    FROZEN                15h

> 2 aliases found [203ms] under stevelee

  source                                      url                                age
  alwaysinmindapi-kxmoebtgsz.now.sh           alwaysinmindapi.now.sh             36s
  alwaysinmind-ovrebfxffj.now.sh              alwaysinmind.now.sh                15h

The alias did seem to take longer than usual though.

[SteveALee]

@matheuss ^^^

[matheuss]

@SteveALee great to know 🎉

@spences10 yes, please share the full output of now && now alias (or the equivalent that you're using) and now ls | head -6 🕵️

[spences10]

Ok cool!

Here are the scripts I'm using in Travis:

package.json

"build": "react-scripts build && mv build c-counter",
"deploy:now": "now c-counter/ --token $NOW_TOKEN",
"alias:dev": "now alias c-counter.now.sh --token $NOW_TOKEN",
"alias:prod":
  "now alias c-counter.scottspence.me --token $NOW_TOKEN",
"cleanup": "now rm c-counter --safe --yes --token $NOW_TOKEN",

.travis.yml

sudo: false

language: node_js

cache:
  directories:
    - node_modules

notifications:
  email: false

node_js:
  - '9'

before_script:
  - npm prune

script:
  - npm i -g now@canary
  - npm run build
  - npm run deploy:now
  - now ls | head -6

after_script: 
  - if [ "$TRAVIS_BRANCH" = "development" ]; then npm run alias:dev; fi 
  - if [ "$TRAVIS_BRANCH" = "master" ]; then npm run alias:prod; fi
  - if [ "$TRAVIS_BRANCH" = "master" ]; then npm run release; fi
  - npm run cleanup

[spences10]

Guessing I need to add --token $NOW_TOKEN to my Travis script?

0.45s$ now ls | head -6
> Error! No existing credentials found. Please run "now login" or pass "--token"

[spences10]

0.31s$ now ls | head -6 --token $NOW_TOKEN
head: unrecognized option '--token'
Try 'head --help' for more information.
> Error! No existing credentials found. Please run "now login" or pass "--token"
> More details: https://err.sh/now-cli/no-credentials-found

[spences10]

It's ok I added the token to now, now instead of head 🙃

- now ls --token $NOW_TOKEN| head -6

[spences10]

@matheuss [Gist] of the Travis output 👍

[codemzy]

@matheuss I can confirm it's working again for me. Thanks for the fix. 👍

[SteveALee]

is there any reason I should not close this now? Thanks for the awesome support!

[playerx]

I'm still facing the same issue:

Green:dist ez$ now && now alias
> Deploying ~/Github/scp-ui/dist under ezeki
> Using Node.js 8.9.4 (default)
> Ready! https://scp-ui-xgixbeuqlv.now.sh (copied to clipboard) [5s]
> Success! Deployment ready!
> Initializing…
> Assigning alias scp-dev to deployment...
> Error! Unknown error: Error: The deployment is not ready yet
Error: The deployment is not ready yet
    at module.exports.Now.createAlias.alias.retry (/snapshot/now-cli/dist/now.js:3986:21)
    at <anonymous>
    at process._tickCallback (internal/process/next_tick.js:188:7)
Green:dist ez$ 

Green:dist ez$ now -v
10.0.0



md5-9ddb3d3d59feba636800d91ebd9d62d3



{
  "name": "scp-ui",
  "alias": "scp-dev"
}

[gkiely]

I just started getting this issue. It used to exit after deploying, now after running now it sits for about 60s then says:

> Deployment complete!
> Running in sfo1