Select Topic Area
General
Body
I have had declining downloads speeds from all github project over the past few months. My ISP is Optimum with the account located in Texas. I have a 300Mbs up/25Mbs down connection with fixe IPv4 address. The download speeds I get from github a basically dial-up speeds of less than 100Kbps.
For example, the last download of mongodb 4.4.29 source took nearly 10 minutes to download 50MB, e.g.
-> Downloading r4.4.29.tar.gz...
% Total % Received % Xferd Average Speed Time Time Time Current
Dload Upload Total Spent Left Speed
0 0 0 0 0 0 0 0 --:--:-- --:--:-- --:--:-- 0
100 49.2M 0 49.2M 0 0 90525 0 --:--:-- 0:09:30 --:--:-- 149k
The individual hop-times seem okay in traceroute between my computer and github.com, but there are 30 hops listed, e.g.
# traceroute github.com
traceroute to github.com (140.82.114.4), 30 hops max, 60 byte packets
1 _gateway (192.168.6.11) 0.751 ms 1.021 ms 0.700 ms
2 * * *
3 173.219.152.4 (173.219.152.4) 13.881 ms 13.866 ms 14.342 ms
4 173.219.197.86 (173.219.197.86) 17.576 ms 17.693 ms 17.549 ms
5 dls-b23-link.ip.twelve99.net (213.248.88.228) 16.846 ms 16.832 ms 19.765 ms
6 dls-bb1-link.ip.twelve99.net (62.115.136.118) 20.769 ms * *
7 * * *
8 * atl-bb1-link.ip.twelve99.net (62.115.137.54) 34.364 ms 35.094 ms
9 rest-bb1-link.ip.twelve99.net (62.115.138.70) 53.207 ms 54.298 ms 53.931 ms
10 rest-b2-link.ip.twelve99.net (62.115.123.41) 53.472 ms 52.949 ms 50.715 ms
11 github-ic-368832.ip.twelve99-cust.net (213.248.67.47) 64.777 ms 50.298 ms 53.864 ms
12 * * *
13 * * *
14 * * *
15 * * *
16 * * *
17 * * *
18 * * *
19 * * *
20 * * *
21 * * *
22 * * *
23 * * *
24 * * *
25 * * *
26 * * *
27 * * *
28 * * *
29 * * *
30 * * *
Is there a general problem with the download bandwidth available for github users or does this point to a technical issue somewhere in the hops between Texas and wherever the github servers are located?
Is this a problem? I had ignore this for long time. Please anyone help to solve the issues.
0 replies
It looks like traffic to github is being routed through twelve99 which appears to be a company based in Sweden. Not sure where the traffic is actually being routed to, but since this was posted 2 days ago and it's still a problem my confidence in whoever is running network operations at github is slightly diminished at this point. I just noticed it this morning trying to download a file that would ordinarily take a minute or two and now it's showing 3 hours.
I tried to find a place to submit a ticket to anything remotely resembling network operations and couldn't find anything so I ended up here. Oh well. I guess we just wait until someone important or connected enough notices and gets it resolved.
0 replies
0 replies
I'm in Poland, ISP is UPC/Play fiber, and similar issue. Download speeds are about 0.5 MiB/s, while for other websites are tens of MiB/s. Downloading bigger things is a problem because it seems there's some timeout on github side - after few minutes download link seems to expire! What is going on?
1 reply
I never had any issues, even using the GitHub Desktop cloning repos and maybe you can try GitHub Desktop too, to see if there is any difference,
also, have you tried downloading it as .zip from GitHub, to again, just test the behavior ? it's in the Code tab
and likely the issues are :
- Network routing issues – Traffic to GitHub may be taking inefficient paths (e.g., routed via twelve99 in Sweden), causing high latency and slow speeds.
- ISP bandwidth throttling or congestion – Your provider might limit speeds during large file transfers or at peak times.
- Temporary regional GitHub CDN slowdown – GitHub’s content delivery network could be experiencing performance issues in your area.
- Local network limitations – Router, firewall, or Wi-Fi bottlenecks affecting sustained transfer speeds.
- Unusually large repository size – Large file history or binaries in the repo can make cloning inherently slower.
3 replies
I'm not talking about cloning git repositories, I'm talking about downloading of artifacts from public repos, e.g. compiled elf, exe or msi files. Some of them are few hundred megabytes in size, and downloading them in my location is nightmare (slow, ends with errors, download links expire - can't be resumed and have start from scratch).
It's not my local network, it's not my network provider - when I tried different provider or different laptop -exactly the same problem. It's something very, veeery wrong with GitHub in Warsaw. We have lots of CDNs here, we have Azure datacenter here, looks like something gets routed terribly.
When I run traceroute I get:
➜ ~ traceroute release-assets.githubusercontent.com
traceroute to release-assets.githubusercontent.com (185.199.110.133), 30 hops max, 60 byte packets
...
6 pl-waw09e-ri1_ae2.0.as9141.pl (185.182.244.27) 25.837 ms 29.089 ms *
7 war-b3-link.ip.twelve99.net (62.115.170.122) 29.088 ms 29.547 ms 36.486 ms
8 ffm-bb2-link.ip.twelve99.net (62.115.120.20) 54.914 ms 35.258 ms *
9 ffm-b17-link.ip.twelve99.net (62.115.132.115) 56.009 ms ffm-b11-link.ip.twelve99.net (62.115.124.119) 34.784 ms 39.565 ms
Yes - this goes via this unfortunate twelve99. How to use a different route?
I have similar issues on my end. I have only one Internet Exchange between my ISP and Fastly, where GitHub servers are located.
Many projects are moving to ghcr.io and therefore I am forced to use GitHubs Container Registry. The download speed for the images are roughly between 20-40 MBit/s.
Please see this traceroute - I cant imagine that my peering is an issue here:
Keys: Help Display mode Restart statistics Order of fields quit
Packets Pings
Host Loss% Snt Last Avg Best Wrst StDev
1. 10.0.0.1 0.0% 422 0.1 0.1 0.1 2.1 0.1
2. 100.126.0.244 0.0% 422 3.0 5.2 2.5 65.0 9.2
3. 100.126.0.46 0.0% 422 3.1 5.0 2.5 62.2 8.7
4. 193.203.0.66 0.0% 421 3.6 3.2 2.7 4.3 0.3
5. 185.199.111.154 0.0% 421 3.4 3.4 2.7 9.0 0.5
193.203.0.66 is Fastly Inc. at VIX and 185.199.111.154 belongs to GitHub via Fastly.
I got a 500 MBit/s downlink from my ISP and all the other content is coming with the exact line speed.
This is really frustrating.
1 reply
For me using a VPN near SF, USA worked. Seems like GitHub CDN is .... bad 🤷🏻
0 replies
And once again I'm sitting here cloning a small repo for 2 hours. Thank you Microsoft!
0 replies
My office and my home which are physically close in the same city use the same ISP but at home I have the slow download speeds from GitHub too.
Tried changing DNS but didn't help.
2 replies
I have the same problem in Germany / Berlin. Abysmall slow download speed with my 500 Mbps line. Changing to Norway via VPN fix that problem. Download speeds are jumping back to ~400 - 500 Mbps. Without VPN I'm stuck with 100 kbps.
0 replies
We good @github?
I don't want to have to wait 50 minutes to setup a Github Actions runner... As @minifisch mentioned GitHubs Container Registry is also affected by this.
I'm on ATT Fiber 500/500
Here's a traceroute if that helps...
traceroute to ghcr.io (140.82.112.34), 64 hops max, 40 byte packets 1 192.168.0.1 (192.168.0.1) 1.309 ms 0.550 ms 0.338 ms 2 192.168.1.254 (192.168.1.254) 0.949 ms 0.943 ms 0.924 ms 3 104-186-148-1.lightspeed.rlghnc.sbcglobal.net (104.186.148.1) 1.552 ms 1.534 ms 5.043 ms 4 99.173.77.18 (99.173.77.18) 3.946 ms 3.220 ms 3.886 ms 5 * * * 6 * * * 7 * * * 8 192.205.32.178 (192.205.32.178) 12.323 ms 11.452 ms 13.038 ms 9 ce-0-2-0.a05.asbnva02.us.ce.gin.ntt.net (157.238.226.211) 10.884 ms 12.421 ms 11.856 ms 10 * * * 11 * * * 12 * * * 13 * * * 14 * * * 15 * * * 16 * * * 17 * * * 18 * * * 19 * * * 20 * * * 21 * * * 22 * * * 23 * * * 24 * * * 25 * * *
1 reply
Yep. This looks like an routing issue between you and GitHub's infra. You can see that there are many ICMP package drops – maybe try a proxy?
0 replies
I think I'm running into this exact same problem as described here. This has been increasingly more problematic over the last several months to the point where now, I'm unable to fully download artifacts from github on a semi-regular basis. Good times.
edit
I'll include what I found in my linked issue above. The current rtw89_8852ce driver is apparently garbage and causes weird connection issues. I removed it entirely from the equation and my desktop can transfer from GitHub at a normal speed again (in line with my 1Gbps fiber).
0 replies
Can someone from Micro$oft eventually look at it? In recent months you broke windows (every update is a nightmare), and you broke github (every download from here in Europe in nightmare because goes to Los Angeles)!
Tried traceroute release-assets.githubusercontent.com also today - the same issues as months ago!
0 replies
I'm experiencing this issue now across all self-hosted github runners trying to pull from the github package registry, taking 10 minutes to download before the runner even starts to process the job.
2 replies
This just started for me today. Many of my GitHub actions that would pull from the GHCR are now failing from the job timing out after 10 minutes when downloading. Ridiculous. Not to mention this is affecting our other servers using ArgoCD.
I don't mind there being issues, inevitable part of the complexity of what they provide, but I find it odd that there was no report of degradation whatsoever on their status page yesterday or today. I find quite often people say it to try to get a reaction or reply or "stick it to" the company - but genuinely I think we'll need to move our runner image hosting elsewhere. Created real blockers for us yesterday having every image pull - from both my local machine and from my runner containers - taking 10+ minutes to pull the image down.
I am experiencing this problem for that last six months. Two connections, same ISP, almost same traceroute, but one connection is getting only 1Mbps and the other is getting line speed. Only difference is these two connections have different IP blocks/subnets.
Have a feeling github/azure/MS is throttling/rate-limiting connections based on IPs.
Below is the result of github-debug.com on the problem connection:
2507998 bytes downloaded from github.com at 1.13 Mbps
2507998 bytes downloaded from cloud.githubusercontent.com at 2.73 Mbps
2507998 bytes downloaded from avatars.githubusercontent.com at 42.15 Mbps
2507998 bytes downloaded from github.githubassets.com at 2.28 Mbps
2507998 bytes downloaded from australiaeast.github-debug.com at 3.31 Mbps
2507998 bytes downloaded from brazilsouth.github-debug.com at 0.43 Mbps
2507998 bytes downloaded from centralindia.github-debug.com at 1.96 Mbps
2507998 bytes downloaded from fra.github-debug.com at 0.53 Mbps
2507998 bytes downloaded from iad.github-debug.com at 0.46 Mbps
2507998 bytes downloaded from israelcentral.github-debug.com at 0.46 Mbps
2507998 bytes downloaded from japaneast.github-debug.com at 0.98 Mbps
2507998 bytes downloaded from koreacentral.github-debug.com at 0.98 Mbps
2507998 bytes downloaded from northeurope.github-debug.com at 0.56 Mbps
2507998 bytes downloaded from sea.github-debug.com at 1.03 Mbps
2507998 bytes downloaded from southafricanorth.github-debug.com at 0.64 Mbps
2507998 bytes downloaded from southeastasia.github-debug.com at 7.55 Mbps
2507998 bytes downloaded from swedencentral.github-debug.com at 0.37 Mbps
2507998 bytes downloaded from uaenorth.github-debug.com at 0.52 Mbps
2507998 bytes downloaded from uksouth.github-debug.com at 0.68 Mbps
2507998 bytes downloaded from github-cloud.s3.amazonaws.com at 0.40 Mbps
202.144.152.xxx <-- This is my IP
Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/144.0.0.0 Safari/537.36
29788a8d2749ec40781431f3f32d8bef
FDB1:4D66D:00AA:65AF:69933163
Only avatars.githubusercontent.com works properly.
0 replies
Given the normal latency but extremely low throughput, this looks more like ISP peering congestion or traffic shaping between Optimum and GitHub’s upstream rather than a GitHub-wide issue; tools like Lemwatch can help confirm if it’s route-specific or localized throttling.
0 replies
Normal latency + extremely low throughput + VPN instantly fixing it = peering / BGP path congestion between your ISP ASN and GitHub’s CDN (Fastly/Azure edge).
This is not DNS, not local Wi-Fi, and not a global GitHub outage.
Different IP blocks on the same ISP behaving differently further confirms route selection / traffic shaping at the ASN level.
There is no client-side fix beyond:
- VPN (different ASN)
- Different ISP
- ISP NOC escalation with traceroute + MTR evidence
This is a routing economics issue between networks, not an application-layer GitHub problem.
0 replies
It's happening to me, is github server side that has a download speed cap?
2 replies
250 kB/sec today. This is just ridiculous.
0 replies