4 ms·
User reports indicate problems at Discord
- curiousgal 5y agoSpotify too!
- derkades 5y agoAlso Wikipedia, Spotify, AWS, GCP
- Jamie9912 5y agoWikipedia fine for me
- derkades 5y agoIt works fine for me too, that's just what downdetector said
- gentleman11 5y agoYoutube is working, and I can log into aws fine, but I'm also seeing down detector agree with you. Must be region specific? https://downdetector.com/status/aws-amazon-web-services/ https://downdetector.com/status/aws-amazon-web-services/ https://downdetector.com/status/google-cloud/ https://downdetector.com/status/google-cloud/
- drocer88 5y agoHave the problems the last couple of weeks been tied directly to the Russia/Ukraine war?
- D13Fd 5y agoI don't think so, and Wikipedia is back for me. But that was my first thought too when I suddenly couldn't get to either Wikipedia or Discord. Both are places where Russians might be getting their own outside information.
- krembanan 5y agoGCP must be down.
- Tijdreiziger 5y ago> Update - We are continuing to investigate the issue impacting the API to find root cause. https://discordstatus.com/ https://discordstatus.com/
- mlex 5y agoDiscord's own status page: https://discordstatus.com/ https://discordstatus.com/
- prions 5y agoDiscord's status page now reports a 99.83% uptime for the last 90 days. Crazy how it seems that they have a major outage this frequently.
- bandwidth 5y agoCould be a GCP outage but not widespread enough for them to report it. Perhaps API gateway is facing issues.
- songzme 5y agoI spend alot of time on Discord on multiple servers. With this outage, I couldn't access any of the servers I'm involved in, which gives me some time to look at open source self hosted alternatives: https://revolt.chat/ https://revolt.chat/ - very similar to discord https://rocket.chat/ https://rocket.chat/ - more similar to slack than discord https://fosscord.com/ https://fosscord.com/ - feels a bit incomplete, I had a hard time using their fosscord instance
- gregoriol 5y agoYou could try Matrix/Element
- holler 5y agoAlso check out Sqwok an alternative that's a sort of hybrid HN/Discord. https://sqwok.im/p/SJab4125ejWHMg https://sqwok.im/p/SJab4125ejWHMg edit: This is a site I'm developing that's meant to be be a place for topical conversation about anything interesting. The main site is of course https://sqwok.im https://sqwok.im, feel free to ask me any questions about it, thanks.
- _1100 5y agoI have to say, as someone that clicked this link that I found a few things you might want to consider before plugging again: 1. No disclaimer that you are involved with the site 2. Maybe talk about something other than furries in the room that you link 3. Maybe link to a landing page with more info on how to...do anything other than talk about furries? Just my 2¢.
- holler 5y agoI should have stated that! The conversations can tend to flow in abstract ways, but that's how chat sort of works? The general rule is that it's Ok so long as it comes back on topic, with the option to just create a new post. Will be adding an easier way to do that soon. lmk if you have any questions
- uo21tp5hoyg 5y ago
- l30n4da5 5y agoMight be region-specific. I was just talking with my friend on Discord about 5 minutes ago. Also noted that Spotify seems to be having issues, but I'm listening to music through Spotify right now without issue. Curious what the affected area for this outage is.
- l30n4da5 5y agoupdate: spotify on my phone is unresponsive. desktop still working fine.
- d3nit 5y ago> Oncall Engineering continues to investigate the root cause of this issue. _We have engaged our partners and are preparing contingencies to restore service._ Well, I don't see that often the last sentence, and honestly it doesn't sounds good at all.
- pfist 5y agoLooks like it's starting to come back. Messages are now loading for me in most servers.
- fkfkno 5y ago
- testplzignore 5y agoLatest update from https://discordstatus.com/ https://discordstatus.com/ : > Engineering has root-caused this issue to a Google Cloud component called "Traffic Director" which is responsible for configuring our load balancing layer. In its malfunction, it caused our internal load balancing layer to not have a valid configuration, which caused a loss of availability of the API. Engineering took measures to remediate by moving to alternate proxy configuration that did not use traffic director. It took us a bit to switch to using a different load balancing topology, but we were able to do so to restore service before Traffic Director issues were resolved. And from https://status.cloud.google.com/incidents/LuGcJVjNTeC5Sb9pSJ9o https://status.cloud.google.com/incidents/LuGcJVjNTeC5Sb9pSJ... : > The issue with Traffic Director has been confirmed to be caused by a recent release; the release has been rolled back and customers can now start using Traffic Director. We have identified a probable root cause and will be publishing an Incident Report within the next several days.