localtunnel is offline.

[thomas-jones]

I get this error when I try to execute the command -

tunnel server offline: connect ECONNREFUSED 138.197.63.247:80, retry 1s

[bentonwong]

Getting the same error as above. Localtunnel was working yesterday

[jaeyun0623]

The same problem I found when newly restarting today.

[jojo-]

Same problem here. Localtunnel was working this morning.

[csimpi]

still doesnt work. I cant do my job, I should test webhooks in Vagrant environment. Whats up guys?

[JamesMGreene]

For me, it stopped working relatively shortly after 2018-01-09T20:59:07.320Z.

[ckl8964]

Same problem.

When will be the service resume ?

[cmirallesp]

The same error, please could you give any feedback?

[frolovdev]

+, same error

[Dan-Ross]

@defunctzombie Any updates about the tool?

[chrisns]

Hey. Is there anything we (the community) can do to help make it more resilient or just pitch in putting some coins in the meter to keep the lights on?

[frolovdev]

But now everything is OK

[defunctzombie]

@chrisns I have upgraded the localtunnel server which should help with it being able to handle the load better.

[yuliwna1]

Hey @chrisns, seems like it's down again. Can you please assist?

[chrisns]

@yuliwna1 not my project. you want @defunctzombie.

to that end @defunctzombie are you ready yet to take some contributions towards that upgraded server?

[defunctzombie]

@chrisns I suspect there are some bugs in the localtunnel server itself that contribute to downtime - tho larger servers mitigate this. Currently the server is a 1GB DigitalOcean instances.

[chrisns]

@defunctzombie happy to have a dig in the code for bugs or at least a write a healthcheck to cycle the instance when it breaks if you can share the detail of how its deployed and what breaks with some logs etc

[defunctzombie]

@chrisns Deployment is setup via the method described here: https://github.com/localtunnel/server#deploy - largely [I think] the issue stems from long lived connections (not closing tunnels) and possibly leading to OOM.

The logs themselves are not always helpful; the latest set contain a lot of:

[error] server:60 null == true
AssertionError [ERR_ASSERTION]: null == true
    at Socket.socketOnData (_http_client.js:423:3)
    at Socket.emit (events.js:182:13)
    at Socket.EventEmitter.emit (domain.js:442:20)
    at Socket.Readable.read (_stream_readable.js:489:10)
    at Socket.socketCloseListener (_http_client.js:327:10)
    at Socket.emit (events.js:187:15)
    at Socket.EventEmitter.emit (domain.js:442:20)
    at TCP._handle.close [as _onclose] (net.js:598:12)
{ Error: write EPIPE
    at WriteWrap.afterWrite [as oncomplete] (net.js:835:14) errno: 'EPIPE', code: 'EPIPE', syscall: 'write' }

Which probably means there are more instances of error handling that need to be done in the server code.

[chrisns]

@defunctzombie to clarify, you're deploying with docker?
if so, if I make a PR to the localtunnel/server with a HEALTHCHECK in the dockerfile and you run it with --restart=always
at least then it'll cycle without manual intervention.

ideally if i can give you a docker-compose.yml with a 3.x schema which would allow multiple instances to run, load balancing and more managable memory limits
if you can let me know the docker engine version running, I can tailor to match.

thoughts?

[defunctzombie]

if so, if I make a PR to the localtunnel/server with a HEALTHCHECK in the dockerfile and you run it with --restart=always
at least then it'll cycle without manual intervention.

If docker has this feature now then yes that sounds about right.

ideally if i can give you a docker-compose.yml with a 3.x schema which would allow multiple instances to run, load balancing and more managable memory limits

This won't "just work" due to how the backend is designed. There is no capability for horizontal scaling of the instances.

[chrisns]

@defunctzombie https://github.com/localtunnel/server/pull/98 when you time