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
pub.dev is down (global outage) #4663
Comments
Internal bug: https://buganizer.corp.google.com/issues/183665226 |
No ETA to fix yet. We're figuring out what's gone wrong; we'll update this bug with status as we learn more. |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Is there any back up website that we can point to while pub.dev is down? |
This comment has been minimized.
This comment has been minimized.
Did not work. |
Updates will be posted at the top of this bug. To make it easier to follow along, please don't post "me too" or +1 comments. |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Kpi is gone this month |
I provide the following solutions, If you can't pack it, please try as follows.
export PUB_HOSTED_URL=~/.pub-cache/hosted/pub.dev
export PUB_HOSTED_URL=~/.pub-cache/hosted/pub.flutter-io.cn Running "flutter pub get" in example... 0.7s Please see below for new solutions #4663 (comment) |
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
This comment has been minimized.
Saw this above. You can try this. If you can access pub.flutter-io.cn, then you can try to use this URL as a temporary solution. export PUB_HOSTED_URL=https://pub.flutter-io.cn ref: https://flutter.dev/community/china doesnt work for me though. goodluck |
the "pub.flutter-io.cn" is accessible |
This comment has been minimized.
This comment has been minimized.
uuuuuuuppppppp |
@timsneath What was the source of the problem? |
pub.dev seems to back, but pub get is aint working |
👍 |
Thank you all. I've posted a quick update at the top of this bug, but in summary services should now be resolved. We've identified the root cause and increased the quota as a short-term measure until we rollback the offending code. |
@timsneath is there any status page for the pub.dev. like https://status.cloud.google.com for GCP? |
Can't speak to the status page yet; we'll figure out the right mitigations during the post mortem. I'm not sure it would have helped us much: the issue page seemed fairly effective to communicate status. But interested to hear from others. Yes, we'll share the post-mortem summary. It should make for fun reading :) We operate a blameless post-mortem policy at Google; it's all about learning lessons rather than finding scapegoats. Any failure is a system failure, and we try and learn how we can address the system causes. |
pub.dev is now accessible from Vietnam. |
Seriously, why didn't you cached YouTube calls for some period of time (eg: release cache a few times a day) in the beginning? |
We do cache them, here is the related code with history: However, once the fetched failed with the quota limit, the error propagated up in the chain - until the isolate was killed, restarted and with the restart we started to fetch it again. We will redesign/refactor this and similar background task so we can make sure such failures will not be propagated in the future. |
Oh I get it. :D Since the previous fetch is failed and the cache was empty the restarted isolate tried to fetch new data again, and it failed & crashed the isolate and gce restarted it then the loop continued... Interesting failure. Thanks for letting us know! |
@timsneath + Google team |
Postmortem is referenced here: https://github.com/flutter/flutter/wiki/Postmortems |
It's still not working. Can't access pub.dev |
@manglide: please open a new issue next time, we don't monitor closed issues. As pub.dev is working for me, it is possible that the problem is at your ISP's side. Please run this script and return back its output: |
Hi @isoos, thanks for your response. The issue is from my local dnsmasq configuration on mac. I have resolved it now and can access pub.dev. Thanks once again. |
Hi, is any problem with this again right now? |
@luis901101 There is an outage right now, we are aware and trying to fix. Also: please don't comment on old threads. |
Is there any public status page for pub.dev? |
@iamchathu no we don't have such a page |
Overview
At approximately 7pm, the pub infrastructure began to respond with a HTTP 502 Site Error message. The site infra is deployed to the
us-central
GCP region. This impacts core Flutter services, specifically:flutter
orpub
commands (e.g.flutter upgrade
orpub upgrade
).Please do not reply to this bug with "me too" or +1 messages, it makes it harder for folk to track. Thanks!
Updates
7:10pm Pacific
We're currently experiencing an outage on pub.dev, which appears to be related to a load balancer issue. We don't have an ETA for a resolution at this time; we're currently working to understand the issue.
7:59pm Pacific
The pub.dev site is still down. We have a Google on-call engineer currently investigating. We have not yet identified a root cause.
8:15pm Pacific
We apologize for the inconvenience. We're seeing load balancer errors and are escalating to the appropriate team. Still no ETA, unfortunately, since we're still haven't determined the root cause.
8:27pm Pacific
We have multiple Google Cloud engineers on-call investigating, but I'm sorry to report that we still don't have a root cause. We'll continue to post updates regularly. Thank you for your patience.
9:00pm Pacific
We are continuing to debug the problem. We have declared a Google escalated outage while we attempt to identify the root cause. Some folk have been successful using the Chinese mirror site at https://pub.flutter-io.cn.
9:20pm Pacific
Again, apologies.
9:35pm Pacific
We are currently exploring the theory that we have exceeded a quota, but that the error didn't show in the log. Paging an oncall team to try and increase the quota to see if this resolves. Again, this really sucks -- we recognize that it's a major inconvenience to you all, and we're feeling sick that we're down. Thank you for being patient with us :(
9:45pm Pacific
We have updated the quota and are resetting the VM instances, to see if we have successfully identified the root cause.
9:51pm Pacific
We are seeing evidence of partially restored service.
9:55pm Pacific
The pub service appears to be fully restored.
10:15pm Pacific
Here's what we think we know at this point in time. At some point within the last day or two, a change was made to the pub.dev landing page that includes a call to the YouTube API. There is a quota limit for YouTube calls that we didn't hit over the last few days, but today we hit it. Confounding the issue, the code was missing exception handling and the logging was inadequate or obfuscated sufficiently that we were unable to immediately spot the problem. The immediate resolution was to raise the quota temporarily to give us time to revert the original change.
At this time we think the issue is resolved, but we'll obviously be monitoring closely. Again, apologies on behalf of the Flutter & Dart teams for the disruption. We take this very seriously, and we will perform a full post-mortem and share the learnings and actions we'll take as a result of this.
The text was updated successfully, but these errors were encountered: