SOLVED Orange circles in Services

robi

graphite
mail
mongo
mysql
postgres
and in some cases..
redis

Several of my cloudrons have after upgrade and browser refresh gone orange, somehow lost the state that the services have actually started and are running, as per their logs.

One can continue to maunally restart services to have them go green, or ..

Alternatively, edit the service configuration and clicking save with no changes for it to recognize it started and turn green.

Can you find the bug?

girish

The orange thing means the services are starting up. Did you happen to capture any service logs when they were orange?

robi

@girish The services are fine. No errors in logs.

The UI seems to not pick up that the services should be green. Until the above actions are taken.

girish

@robi Ah, I see some UI issue. Thanks for reporting, will have to take a look.

d19dotca

I guess you're already aware of the UI issue, but saw this as I was about to report this too. Seems I have the exact same issue reported by @robi in terms of services affected as well.

I noticed in my Services tab today that several are continually in "starting" state even though they're actually operating properly and their logs can confirm it. While it says "starting", the service is actually running and should be green.

Screen Shot 2021-02-14 at 6.56.45 PM.png

robi

@d19dotca thank you for the additional confirmation, data point, perspective and wording

nebulon

if this comes up again for one of you, can you open the webinspector's network tab and see if there are new requests happening to fetch the current service status and if so, can you dump maybe a few examples of the response body for investigation?

tomw

@nebulon I experienced something similar. Currently graphite, mail, mongodb and mysql appear orange. Here's the response for graphite, for example:

{
  "service": {
    "name": "graphite",
    "status": "starting",
    "error": "Error waiting for graphite: ESOCKETTIMEDOUT",
    "healthcheck": null,
    "config": {
      "memoryLimit": 268435456
    }
  }
}

Each time I refresh the page, it seems to be a lucky dip which of those come back as green and which as orange. e.g. a second time everything was green except mail, a third time mysql is orange again.

nebulon

@tomw this actually then sounds like the UI is showing the correct state, however the addons are not reliably and timely responding to the status call.

d19dotca

For what it's worth, I'm currently running on Ubuntu 20.04, not sure if that's making a difference at all since the add-on components would be running a different version I presume (i.e. MySQL, PostgreSQL, MongoDB, etc.), though I guess that doesn't necessarily make sense for the Mail service.

scooke

FWIW, I just checked my Services, and saw that mail and mongodb were orange. So I tweaked the memory just a tiny little bit (adding 124 MB), and the circle almost immediately went green. I don't know what that might add to the search, but if you want a green circle, this seems to be one way to get it!

girish

OK, I have added a fix for the next release to clear the timeout when getting service status. This should hopefully mitigate the issue.

subven

Hmm I have a similar issue on a Cloudron (v6.1.2 and before) where random apps are stuck in "Starting" but apparently working/running as they should. Services however work just fine on this Cloudron.

2021-02-16 20_35_51.png

This state is kinda permanent and restarting the app nor the server does not help. Also seems to be kinda random what app is affected but their numbers seem to grow. My other two Cloudrons are fine ^^

Log output as en example (for Element) is:

Feb 16 20:36:35 box:settings initCache: pre-load settings
Feb 16 20:36:35 box:taskworker Starting task 789. Logs are at /home/yellowtent/platformdata/logs/e0016eed-67a4-45dd-8234-642c8aee11d3/apptask.log
Feb 16 20:36:35 box:tasks 789: {"percent":2,"error":null}
Feb 16 20:36:35 box:apptask chat.heregoesthedomain.com startTask installationState: pending_restart runState: running
Feb 16 20:36:35 box:tasks 789: {"percent":20,"message":"Restarting container"}
Feb 16 20:36:35 http-server stopped.
Feb 16 20:36:36 ==> Starting Riot
Feb 16 20:36:36 box:tasks 789: {"percent":100,"message":"Done"}
Feb 16 20:36:36 box:apptask chat.heregoesthedomain.com updating app with values: {"installationState":"installed","error":null,"health":null}
Feb 16 20:36:36 box:taskworker Task took 1.109 seconds
Feb 16 20:36:36 box:tasks setCompleted - 789: {"result":null,"error":null}
Feb 16 20:36:36 box:tasks 789: {"percent":100,"result":null,"error":null}
Feb 16 20:36:36 Starting up http-server, serving /app/code
Feb 16 20:36:36 Available on:
Feb 16 20:36:36 http://127.0.0.1:3000
Feb 16 20:36:36 http://172.18.16.9:3000
Feb 16 20:36:36 Hit CTRL-C to stop the server
Feb 16 20:36:57 [2021-02-16T19:36:57.389Z] "GET /" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/88.0.4324.150 Safari/537.36"

PS: Sorry, not wanted to hijack this topic. I thought about opening an seperate issue for this but so far it was not that much of a problem and I hoped it would be fixed after the next update ^^

girish

@subven I think the starting issue is something else. If you do systemctl restart box, I think they will all become running again. Can you please try?

subven

@girish said in Orange circles in Services:

@subven I think the starting issue is something else. If you do systemctl restart box, I think they will all become running again. Can you please try?

Yes they are I was confused because even box:apphealthmonitor said the apps are up.

robi

something is getting lost between the checks and the UI representation.

d19dotca

This appears to still be an issue on 6.2.1 by the way, though I seem to get this mostly with Mail now more than the others.

subven

Kinda same for me. Freshly installed server + imported Cloudron 6.1.2 backup. The server the backup is from did not had the problem.

2021-03-07 20_09_28-Window.png

nebulon

@subven interesting that it shows an "active" tooltip. Is this state consistent across page reloads?

subven

@nebulon said in Orange circles in Services:

@subven interesting that it shows an "active" tooltip. Is this state consistent across page reloads?

Mostly yes, red circles vary after reload. Seems to be my local Chrome because it works with FF and on other machines with Chrome. Strg+F5 and deleting the local site data storage does not help. It's fine for me since its only cosmetic.

nebulon

@subven I think this is some client side state race or so. Any information here in behavior can help

subven

@nebulon I solved it. It was caused by one of my chrome addons (Dark Reader)

robi

@subven can you be more specific?

subven

@robi regarding to what? I had a similiar issue than other users (multiple false positive red circles) and I fixed it by disabling a Chrome addon (Dark Reader) that alters page CSS. I can reproduce it by enabling the addon for the dashboard domain.

Dark Reader turned on:
2021-03-08 20_30_16-Window.png

Dark Reader turned off:
2021-03-08 20_30_41-Window.png

robi

@subven thanks, thought perhaps you modified a setting within DR

d19dotca

@subven I think that may be specific to your use-case but it's still good to know it's fixed for you. I'm using Safari and have no extensions installed other than a password manager. Nothing that interferes with the webpage content. I've disabled some settings too as a result though to test further (like Safari's non-tracking features and stuff) but no difference in my case. I also use Mobile Safari where I see the issue and that has zero extensions on it at all.

I could be wrong, but I think the issue started around the time 6.1.2 came out. I'm not convinced this is a browser issue, although I think that's what @nebulon is suggesting if I understand him correctly.

@robi - in case this is indeed a browser / client-side issue (which I don't think it is personally), maybe note which browser you're using and accessing Cloudron on where you see this issue too in case it's Safari for both you and me. I suspect the behaviour @subven has seen is entirely separate from what you and I are reporting, judging by their comment on Feb 16th although admittedly similar in aspects.

nebulon

@d19dotca I also think that maybe that darkmode plugin just changes the timing and does not cause the race to hit badly.

robi

@nebulon for me, this still happens in a non dark mode enabled chrome based browser.

girish

@robi Is this the same server that had the etherpad issue? Because that one seems just fine here. I tried refreshing many times now, always green.

necrevistonnezr

@girish
I, too, have several services from time to time indicating that they are starting though they seem to be running fine - such as mail in this example.

Screenshot_2021-03-09 CKFL.png

nebulon

This is a very curious case. The tooltip of that status indicator is using the exact same property as the color. Both are not updated individually. So far I was not able to see how there could be some states out-of-sync if status is active.

I do see from the screenshots that the restart button also spins though, this indicates in some of those screenshots that the status is indeed starting not active for which the UI then seems to be consistent.

This leads to a different question, if the service is actually running fine, why does the REST api then report starting? From the code, this only happens if the upstream service instance (they live in their own container) fails to report status, while the container being up.

Now to get more information, it would help if some of you, who can reproduce this, could take a look at the box logs at /home/yellowtent/platformdata/logs/box.log while refreshing the dashboard. This should trigger a call to all the service containers and thus should hopefully give us more information. For example the call might just timeout or otherwise fail.

robi

@nebulon After waking up a sleeping tab, I had graphite and mongo in orange. Graphite was spinning.

After reloading the services page, mongo went green. I then edited the graphite service and just saved. It then went green too.

these are the box logs.

2021-03-09T16:50:21.009Z box:translation Languages found: [ 'de.json', 'en.json', 'it.json', 'nl.json', 'zh_Hans.json' ]
2021-03-09T16:50:21.332Z box:apphealthmonitor app health: 21 alive / 1 dead.
2021-03-09T16:50:23.093Z box:shell statusUnbound exec: systemctl is-active unbound
2021-03-09T16:50:23.146Z box:shell statusNginx exec: systemctl is-active nginx
2021-03-09T16:50:23.178Z box:shell statusUnbound (stdout): active

2021-03-09T16:50:23.178Z box:shell statusUnbound (stderr):
2021-03-09T16:50:23.193Z box:shell statusNginx (stdout): active

2021-03-09T16:50:23.193Z box:shell statusNginx (stderr):
2021-03-09T16:50:31.262Z box:apphealthmonitor app health: 21 alive / 1 dead.
2021-03-09T16:50:41.027Z box:apphealthmonitor app health: 21 alive / 1 dead.
2021-03-09T16:50:50.875Z box:apphealthmonitor app health: 21 alive / 1 dead.
2021-03-09T16:50:59.971Z box:addons updateServiceConfig: graphite {"memoryLimit":268435456}
2021-03-09T16:51:00.038Z box:shell update(graphite) spawn: /usr/bin/docker update --memory 241172480 --memory-swap 268435456 graphite
2021-03-09T16:51:00.735Z box:apphealthmonitor app health: 21 alive / 1 dead.
2021-03-09T16:51:01.190Z box:shell update(graphite) (stdout): graphite

2021-03-09T16:51:10.840Z box:apphealthmonitor app health: 21 alive / 1 dead.

d19dotca

@nebulon If I refresh the page several time in a row, it seems to get them in sync again and they all show green. There were 3 orange when I first loaded it and even after about two refreshes. It seemed to take around 3 or more refreshes to get it all green again.

For what it's worth, the services are definitely running despite it showing the restart icon spinning. And that's the issue for me... if I were to trust the page when I first load it, it'd imply my mail server is never running, lol, which of course it is when I check the logs for the mail server (and the fact I've had no customer complaints on mail server not working).

Here's the logs and what I see after refreshing several times to when it became green again. Hopefully it helps. I can't make heads or tails of it myself, haha, mostly because I was expecting to see something like statusMail but don't see that anywhere, only statusNginx and statusUnbound.

2021-03-09T18:43:25.216Z box:shell statusUnbound exec: systemctl is-active unbound
2021-03-09T18:43:25.281Z box:shell statusUnbound (stdout): active

2021-03-09T18:43:25.284Z box:shell statusUnbound (stderr): 
2021-03-09T18:43:25.285Z box:shell statusNginx exec: systemctl is-active nginx
2021-03-09T18:43:25.381Z box:shell statusNginx (stdout): active

2021-03-09T18:43:25.384Z box:shell statusNginx (stderr): 
2021-03-09T18:43:30.373Z box:apphealthmonitor app health: 27 alive / 0 dead.
2021-03-09T18:43:40.235Z box:apphealthmonitor app health: 27 alive / 0 dead.
2021-03-09T18:43:50.289Z box:apphealthmonitor app health: 27 alive / 0 dead.
2021-03-09T18:43:58.377Z box:shell statusUnbound exec: systemctl is-active unbound
2021-03-09T18:43:58.389Z box:shell statusNginx exec: systemctl is-active nginx
2021-03-09T18:43:58.422Z box:shell statusUnbound (stdout): active

2021-03-09T18:43:58.422Z box:shell statusUnbound (stderr): 
2021-03-09T18:43:58.425Z box:shell statusNginx (stdout): active

2021-03-09T18:43:58.425Z box:shell statusNginx (stderr): 
2021-03-09T18:44:00.894Z box:apphealthmonitor app health: 27 alive / 0 dead.
2021-03-09T18:44:05.161Z box:shell statusUnbound exec: systemctl is-active unbound
2021-03-09T18:44:05.212Z box:shell statusNginx exec: systemctl is-active nginx
2021-03-09T18:44:05.234Z box:shell statusUnbound (stdout): active

2021-03-09T18:44:05.234Z box:shell statusUnbound (stderr): 
2021-03-09T18:44:05.270Z box:shell statusNginx (stdout): active

2021-03-09T18:44:05.270Z box:shell statusNginx (stderr): 
2021-03-09T18:44:08.914Z box:shell statusUnbound exec: systemctl is-active unbound
2021-03-09T18:44:08.936Z box:shell statusNginx exec: systemctl is-active nginx
2021-03-09T18:44:08.958Z box:shell statusUnbound (stdout): active

2021-03-09T18:44:08.958Z box:shell statusUnbound (stderr): 
2021-03-09T18:44:08.961Z box:shell statusNginx (stdout): active

2021-03-09T18:44:08.961Z box:shell statusNginx (stderr): 
2021-03-09T18:44:10.812Z box:shell statusUnbound exec: systemctl is-active unbound
2021-03-09T18:44:10.826Z box:shell statusNginx exec: systemctl is-active nginx
2021-03-09T18:44:10.858Z box:shell statusUnbound (stdout): active

2021-03-09T18:44:10.859Z box:shell statusUnbound (stderr): 
2021-03-09T18:44:10.869Z box:shell statusNginx (stdout): active

2021-03-09T18:44:10.869Z box:shell statusNginx (stderr):

And when I refresh the page again just now (after it was all green) it went orange again. Here are the logs when I refreshed and it went orange again...

2021-03-09T18:49:30.135Z box:apphealthmonitor app health: 27 alive / 0 dead.
2021-03-09T18:49:31.977Z box:shell statusUnbound exec: systemctl is-active unbound
2021-03-09T18:49:32.004Z box:shell statusNginx exec: systemctl is-active nginx
2021-03-09T18:49:32.034Z box:shell statusUnbound (stdout): active

2021-03-09T18:49:32.034Z box:shell statusUnbound (stderr): 
2021-03-09T18:49:32.069Z box:shell statusNginx (stdout): active

2021-03-09T18:49:32.070Z box:shell statusNginx (stderr): 
2021-03-09T18:49:40.225Z box:apphealthmonitor app health: 27 alive / 0 dead.

d19dotca

@nebulon If it helps, I took a screen recording of the Services page with the logs beside it, and refreshed several times. This time I wasn't able to get them all green at one time no matter how many refreshes were done or how quickly they were done.

You'll also see in the video that it's kind of random which ones are orange. Generally it's between a few of them, but every refresh isn't the same outcome. Out of 5 possibly orange for example, it'll be maybe 2 orange the next refresh, then 3 the next refresh, then 1 the next one, etc.

Here's the download link for the video, hopefully this helps:
https://files.d19.ca/Orange%20circles%20movie%20refreshing%20with%20logs.mov

One thing (and why I think this isn't a browser issue and is a Cloudron issue) is that the ones that are orange always seem to be the same. For example, I've personally never noticed the sftp one ever being orange nor unbound nor nginx. It's only ever mail, mongodb, mysql, postgresql in my case (I may be missing one).

girish

@d19dotca Are you able to drop us an email at support@ with support access? Looks like we can figure this one lot quicker that way.

d19dotca

@girish Certainly. I'll send an email in a few moments. Thanks for looking into this.

Edit: Sent via the form on the Support page inside Cloudron server, I made sure you have SSH access too.

girish

@d19dotca yes, it's not nice if that view is not reliable. worth fixing.

girish

Found the problem. The timeouts for the healthcheck query was low. https://git.cloudron.io/cloudron/box/-/commit/9650a55c8500ad6c18a175289556c09967c7252a . This also fixes the curious case of graphite showing the yellow indicator in many cloudrons (which has been reported many times in the forum already)!

d19dotca

@girish Thanks for taking care of that - I really appreciate it!

robi

Amazing!

And so we all get a confidence boost with a simple timeout fix!

Thanks @girish & everyone who contributed on this thread!

Love you all

Cloudron makes it easy to run web apps like WordPress, Nextcloud, GitLab on your server. Find out more or install now.

SOLVED Orange circles in Services