gitpod
gitpod copied to clipboard
gitpod-io
Epic: more regional workspace clusters
Summary
Add workspace-clusters to more regions, we currently use two (europe-west and us-west).
- [ ] Identity regions needing workspace-clusters
Context
Customers with high latency are less likely to adopt or continue using Gitpod.
Depending on where you live in the world, latency can be so high that adopting Gitpod is not possible due to latency as a result of distance from the existing EU or US clusters. Further, adoption may be possible, but retention may not because its "just" not quite fast enough. In other words, fast enough to 'check out' but not 'consistently use'.
Test your own latency to regions:
- Test your latency across GCP regions, load the page, wait 10s, scroll to bottom, click stop, then sort by mean
- What is your latency like in europe-west, us-west, asia-southeast, and asia-south? A mean > 100ms is not ideal.
Additional context: Prior issue Related epic from Meta to simplify multi-meta
Value
Users will be more likely to adopt and continue using Gitpod saas because the speed is conducive to a good user experience. For example, as you type quickly, you expect to see those characters instantly.
Acceptance criteria
New workspace clusters are
- Functional in new target regions
- Connected to a meta cluster in the new target region (this will ensure a good dashboard experience)
- Observed by our monitoring
Measurement
Adoption and retention for customers location in regions who currently lack workspace clusters will increase.
Growth Area
Adoption - new customers will be more likely to try Gitpod because its not slow on first use Retention - customers who adopt Gitpod will continue using because it is just as fast as applications on their local machine
Hypothesis
If we add workspace-clusters to new regions, users will be more likely to adopt and continue using Gitpod in those regions.
Complexities
- Do we need to deploy meta to these regions? For adoption, no, but for retention, probably. Otherwise dashboard usage will not be ideal.
- Short term, if we wanted to run an experiment, perhaps we could deploy to another region and connect to meta in EU or US. @csweichel wdyt?
- How will monitoring work for workspace-clusters in new regions?
- @meysholdt @mads-hartmann can you help set our expectations? It looks like we have a single instance of our monitoring in europe-west, therefore I assume we would not need additional instances of monitoring in new regions to support new workspace-clusters running?
- Depending on our forecasted usage for the new region, and the cost for nodes in other regions, we may want to consider using a size other than our default.
Related
- https://github.com/gitpod-io/gitpod/issues/1496
Identity regions needing workspace-clusters
It's clear that we want to have a Singapore cluster (asia-southeast1) to fix the current issue we have in APAC, and would include only that one in the spirit of MVC.
Given our focus on reducing latency to values below 100ms, it might be relevant to consider South America but I wouldn't attribute the same level of urgency given the feedback from the community, data on usage, and resource costs. Also worth noting that there's no GCP region in Africa.
Related issues:
- https://github.com/gitpod-io/gitpod/issues/5534
- https://github.com/gitpod-io/gitpod/issues/6139
- https://github.com/gitpod-io/customers/issues/6
Related resources:
+1000 to all the things said above. I got a self-hosted gitpod instance in asia-east1 and the latency improvements are significant, and make Gitpod seem a lot more fluid.
Ping times for SAAS vs Self-hosted (in `asia-east1`)
on ⛵ gke_self-hosted-tarun_asia-east1_gitpod ~ on ☁️ [email protected]
❯ ping -c 10 gitpodio-gitpod-1osgnsetjzz.ws.tarun.gitpod-self-hosted.com
PING gitpodio-gitpod-1osgnsetjzz.ws.tarun.gitpod-self-hosted.com (35.229.210.95) 56(84) bytes of data.
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=1 ttl=110 time=92.3 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=2 ttl=110 time=92.7 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=3 ttl=110 time=91.1 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=4 ttl=110 time=90.1 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=5 ttl=110 time=93.6 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=6 ttl=110 time=91.8 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=7 ttl=110 time=89.8 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=8 ttl=110 time=89.9 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=9 ttl=110 time=103 ms
64 bytes from 95.210.229.35.bc.googleusercontent.com (35.229.210.95): icmp_seq=10 ttl=110 time=90.2 ms
--- gitpodio-gitpod-1osgnsetjzz.ws.tarun.gitpod-self-hosted.com ping statistics ---
10 packets transmitted, 10 received, 0% packet loss, time 9011ms
rtt min/avg/max/mdev = 89.848/92.431/102.794/3.666 ms
on ⛵ gke_self-hosted-tarun_asia-east1_gitpod ~ on ☁️ [email protected]
❯ ping -c 10 gitpodio-gitpod-itpc6y1mcb6.ws-us30.gitpod.io
PING gitpodio-gitpod-itpc6y1mcb6.ws-us30.gitpod.io (34.105.101.238) 56(84) bytes of data.
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=1 ttl=105 time=229 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=2 ttl=105 time=253 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=3 ttl=105 time=202 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=4 ttl=105 time=297 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=5 ttl=105 time=203 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=6 ttl=105 time=202 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=7 ttl=105 time=274 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=8 ttl=105 time=290 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=9 ttl=105 time=216 ms
64 bytes from 238.101.105.34.bc.googleusercontent.com (34.105.101.238): icmp_seq=10 ttl=105 time=203 ms
--- gitpodio-gitpod-itpc6y1mcb6.ws-us30.gitpod.io ping statistics ---
10 packets transmitted, 10 received, 0% packet loss, time 9000ms
rtt min/avg/max/mdev = 201.722/236.953/297.339/36.487 ms
A min of 201 ms is definitely not ideal! :)
For me in Australia the only regions from GCP that get under 100ms were the Australian regions which for Data Sovereignty issues I would love as well. Otherwise the next best based on mean is asia-southeast2
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
![stale[bot] avatar](https://avatars.githubusercontent.com/in/1724?v=4 "Published by a pub.dev verified publisher"){kind=small}
Hey @ghuntley! This issue is open 🙌 Unless I'm misunderstanding what you mean something else? 🦩
There seems to be some confusion relating to the closing of the following issues:
- https://github.com/gitpod-io/gitpod/issues/5534
- https://github.com/gitpod-io/gitpod/issues/6139
Please note: This is not a an update of the status of this issue (closing or opening). The reason for the consolidation is that we now have a proliferation of different regional issues being raised by users. It doesn't make sense for users to directly raising issues about specific required regions, as when it comes to rolling out new regions we'll need to review:
- Costings
- Latencies
- Current + predicted future usage of Gitpod
...and many other factors.
It makes sense for this conversation to happen in a single issue, which is this one. Please continue to share any thoughts or feedback. We'll look to update the description soon as a consolidation of current thinking, as there are now multiple region related issues with many comments.
Thanks all 🙏
+1 forsouthafricanorth or af-south 😬 Would love to dodge the 190+ ping without needing to self-host
Can't believe there's still no progress on this issue after nearly a year and a half, I've heard a gitpod ex-employee complain about high latency pain within your team (in Asia Pacific or Oceania) . What I really can't understand is that this requirement is to simply deploy an additional k8s cluster in another data center of GCP, but the gitpod team seems to have no motivation to do this at all. Is it because the number of users in APAC is too small to be profitable？
Can't believe there's still no progress on this issue after nearly a year and a half, I've heard a gitpod ex-employee complain about high latency pain within your team (in Asia Pacific or Oceania) . What I really can't understand is that this requirement is to simply deploy an additional k8s cluster in another data center of GCP, but the gitpod team seems to have no motivation to do this at all. Is it because the number of users in APAC is too small to be profitable？
Very good. I support what you said.
Can't believe there's still no progress on this issue after nearly a year and a half, I've heard a gitpod ex-employee complain about high latency pain within your team (in Asia Pacific or Oceania) . What I really can't understand is that this requirement is to simply deploy an additional k8s cluster in another data center of GCP, but the gitpod team seems to have no motivation to do this at all. Is it because the number of users in APAC is too small to be profitable？
This is exactly right. I switched back to mostly local development because gitpod is just too slow. It's actually been two years since this was original promised.
Although I bought a one-year license, I didn't use it for an hour. There will be no renewal later.
Thank you @xzmeng, @bbhxwl and @smacintyre for caring so much about Gitpod, and again raising this issue with us—we really do hear your feedback, and we have been discussing internally this issue / feature and how we can make it happen. It’s not easy to see the work that Gitpod is doing behind-the-scenes to progress with issues like this one, but believe us when we say that we’re working towards it. Sadly, it isn’t as straightforward as simply deploying a new cluster. If it were we probably would have shipped this feature already to the community. We really do appreciate your patience, and hopefully we can provide another update soon. Thanks for caring so much about Gitpod and the experience! 🙌
I am in China and hope to have Hong Kong, Taiwan, Japan or Singapore
I prefer Taiwan because it just at the middle of east Asia
and even GitHub codespaces have Asia server...
Best latency for my team is australia-southeast1 at <15ms.
Current latency is >150ms for us-west.. couple with that getting test data close enough to make things workable.
Singapore for the record is >100ms here in Sydney.
Australian deployment option would be great.
Hey, is there any updates on the ability to chose a region in asia? It's been years in the making now...
Hey guys, any news? That's been years now... We are abandoning gitpod because of this, quite sad.
Hey guys, any news? That's been years now... We are abandoning gitpod because of this, quite sad.
I have already given up, and I also need Asia
@LouAdrienCEL GitHub Codespaces offer regional servers... Just sayin
@electrolyte-orb thanks for the feedback! Yes aware of codespace (that's actually what we are moving to), but they are tightly coupled with github, while we are on bitbucket. So would have prefered a tool that is more platform agnostic!
Cheers
@LouAdrienCEL not a big deal... Create a custom dockerfile that automatically checks in bitbucket repo on boot