mirror of
https://github.com/fosrl/pangolin.git
synced 2026-05-07 21:30:36 -05:00
[GH-ISSUE #1724] Unable to create Sites #6788
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @Lokowitz on GitHub (Oct 21, 2025).
Original GitHub issue: https://github.com/fosrl/pangolin/issues/1724
Describe the Bug
Notices today that i am not able to create a new site anymore.
NewtandBasic WGare grayed out not selectable, justLocal.So sounds like a connection problem to Gerbil. But my existing Sites and Resources are working still fine.
Maybe something related to the Gerbil default port change?
Only interesting from the logs, Gerbil is spamming all the time:
Pangolin looks good:
Environment
To Reproduce
Not sure if this can be reproduced easily. My Setup is just regulary upgraded since beta phase.
Expected Behavior
Able to add new Sites
@baldemar-wuda commented on GitHub (Oct 21, 2025):
Same thing for me.
@Lokowitz commented on GitHub (Oct 21, 2025):
And I tried to go back to older versions:
Pangolin 1.10.3 and 1.11.0
Gerbil 1.2.0 and 1.2.1
But no combination solved the issue.
@oschwartz10612 commented on GitHub (Oct 21, 2025):
e447549de1should fix this when 1.11.1 isreleased again.
@oschwartz10612 commented on GitHub (Oct 21, 2025):
Could you check the latest 1.11.1 release?
@Tuphal commented on GitHub (Oct 22, 2025):
Still not possible to add a new tunnel.
@Lokowitz commented on GitHub (Oct 22, 2025):
Also for me not working with the new image.
@Lokowitz commented on GitHub (Oct 22, 2025):
Gerbil spamming is caused through olm holepunch, I turned it off for now.
Should be related to my CGNAT IPv4, hope support for IPv6 is coming :)
@oschwartz10612 commented on GitHub (Oct 22, 2025):
Hum this is caused when there is not an exit node in the database to
assign a site to. This is usually when there is no gerbil registered but
the update should not have caused this but who knows...
Couple of questions:
Are you on enterprise or community
If you restart the whole stack
docker compose downanddocker compose upto get gerbil to register does it work?If not when this occurs could you check the console logs and network
tab in the browser debugger. Looking for an error message or a 500/400
error call to pick-site-defaults in the network tab. LMK if this is the
case.
If you have a backup database, roll back to that and then upgrade
again and see if it occurs. Make sure it still works on your old database.
Run the following commands
sudo apt install sqlite3
sqlite3 config/db/db.sqlite "SELECT * FROM exitNodes"
And then let me know what the output is
@Tuphal commented on GitHub (Oct 22, 2025):
Community edition
Gebril is connected. The first site is working without issues
This is pointing to a right direction..
The API shows:
{"data":null,"success":false,"error":true,"message":"No available subnets","status":500,"stack":null}I could not find any config variable to change the subnet. The org-subnet should be enough big for a second site.
Edit: I used DB Browser for my downloaded database. Here is the exitNodes entry:
Edit 2: Should the subnet from the site not be part of the adress?
Subnet 100.89.128.8/30Adress 100.89.128.1/24@Lokowitz commented on GitHub (Oct 23, 2025):
I got deeper in the database and found some inconsistencies in ip addresses and subnets.
I updated the database and now It is working again, maybe it was caused through a wrong migration task and/or outdated configs.
1. Check
config/config.ymlI still had the subnet group configured and this is not required anymore -> so remove
2. Check database table orgs
I have 2 orgs, one really old one (since beta phase) and one is maybe 2-3 months old
The old org has subnet
100.90.128.0/24and the newer one has100.89.138.0/20, the new default should be100.90.128.0/24, but is matching the value inconfig.yml.Line 3 is just a test org to get the default values.
3. Check database table sites
Subnet for my first site (local) is set to
0.0.0.0/0a new created is getting0.0.0.0/32.The other newt sites have
100.89.128.4/30,100.89.128.8/30,100.89.128.12/30what is looking good so far.SiteId 1, 4, 5 is associated to org 1, SiteId 9 to org 2 and SiteId 12 to org 3 (test).
And now you see that address of SiteId 9 is not matching the subnet of org 2.
Steps to fix it
100.90.128.0/240.0.0.0/32NULL100.90.128.x/24to match the org subnet (increase x for each site in the same org)@Tuphal commented on GitHub (Oct 24, 2025):
Thank you ver much @Lokowitz for the steps.
These worked perfectly !!!
In my case I only had to follow step 2 and 3
@elcajon commented on GitHub (Oct 28, 2025):
Also run into this today.
Changing the local subnet from /0 to /32 fixed it (address was already empty)
Database migration table content for reference:
@intersectRaven commented on GitHub (Nov 4, 2025):
I experienced this just yesterday and this fixed it.
@oschwartz10612 commented on GitHub (Nov 4, 2025):
Thanks so much @Lokowitz that was indeed an issue on an older version. I double checked and this should not occur on latest versions.
@github-actions[bot] commented on GitHub (Nov 19, 2025):
This issue has been automatically marked as stale due to 14 days of inactivity. It will be closed in 14 days if no further activity occurs.
@ALERTua commented on GitHub (Dec 17, 2025):
The issue persists on Pangolin v1.13.1.
The workaround still helps.
@cpuks commented on GitHub (Dec 19, 2025):
Yeah v1.13.1 still has this bug and fix works
@dr-jeleise commented on GitHub (Dec 28, 2025):
May I ask a question for my better understanding of the architecture.
In the famous example of Lokowitz he shows an example of the table sites. Why ist the subnet of the entries (100.89.128.x) different from the corresponding adresses (100.90.128.x)? How do the 89 and 90 match?
@nlsrchtr commented on GitHub (Feb 24, 2026):
Hi @oschwartz10612,
many thanks for your work on this wonderful piece of software. It helped me a lot so far!
I was hit by this error message today as well, but the details are bit different. I just want to post here, so others might see it as well, since it's related. My highest subnet is "100.89.128.252/30" and Pangolin doesn't allow me to create new subnets anymore. (Yes, there are around 60 sites connected).
What is a good way forward to extend the available subnets for Pangolin, so it doesn't get "confused" and the connected sites are still not able to talk to each other?
Thanks for any help on this!
Regards,
@oschwartz10612 commented on GitHub (Feb 25, 2026):
@nlsrchtr You can open up the subnet on the gerbil config: https://docs.pangolin.net/self-host/advanced/config-file#param-subnet-group
@nlsrchtr commented on GitHub (Feb 26, 2026):
Hi @oschwartz10612,
thanks for coming back on this. Since there is nothing defined in my config, I assume the default value is the "Example" value (
10.0.0.0/8) from the documentation, right?If I see this correctly, this subnet defines IP ranges from
10.0.0.1up to10.255.255.254which is more than enough for me.But Pangolin seems to automatically assign subnets only with
100.89.128.X/30and also blocks the creation of new sites in the UI as well.Is there anything I'm missing here?
Thank you for looking into this.
@oschwartz10612 commented on GitHub (Feb 26, 2026):
We should update the documentation. The default is 100.89.137.0/20
@nlsrchtr commented on GitHub (Mar 2, 2026):
Thanks for the information. With this netmask there are 4094 hosts available. The IPs range is from
100.89.128.1upto100.89.143.254.I have 60 sites connected and only the
100.89.128.Xis "full", but wouldn't Pangolin take another IP address from the range above? An IP from the next range100.89.129.1for example?It seems to me, I'm missing something here? Could you enlighten me once more, @oschwartz10612?
@nlsrchtr commented on GitHub (Mar 9, 2026):
Hi @oschwartz10612,
just wanted to ping you on this issue here again, since I'm still wondering if this situation is a bug or a limitation in my understanding of the subnets: If the default defined subnet of
100.89.137.0/20is offering 4094 hosts, why is Pangolin not using something outside of100.89.137.0 - 100.89.137.255, when that part of the IP range is already used?Do you need any debugging, logs, configs or something else to dive into this deeper?
@oschwartz10612 commented on GitHub (Mar 26, 2026):
@nlsrchtr perhaps you have a different default. We may have had a /24 in there for some time which could be causing your issue. Could you try to set the subnet group to a larger subnet and restart gerbil and pangolin and test again.
Before doing that you could try to do the following commands to see what the actual value is that gerbil is assigning sites from:
And let me know the output
@nlsrchtr commented on GitHub (Mar 26, 2026):
Hi @oschwartz10612,
thanks for getting back to me! You are right - in the output there is a /24 net defined:
100.89.128.1/241|Exit Node XXXX|100.89.128.1/24|XXXX|XXXX|51820|http://gerbil:3003||1||gerbil|What would be a stable way out of this situation? Is it just setting the subnet group to
100.89.137.0/20and hoping all site will reconnect and everything works out? That would be great.Regards,
@oschwartz10612 commented on GitHub (Mar 26, 2026):
I would set it to 100.89.128.1/20 and you should be good I think! That way the existing sites are still included in the range.
@nlsrchtr commented on GitHub (Mar 26, 2026):
I'm sorry, but this didn't worked out. I set:
and even rebooted the whole installation. But after that, I was still not able to add new sites. Looking into the SQlite DB, the entry is still
100.89.128.1/24. Do I need to update that manually in the database?@oschwartz10612 commented on GitHub (Mar 26, 2026):
Ahh this was my mistake. It does not copy it in. Apologies. Yes, you may need to set this manually.
If this does not fix it then we could be having issues with the subnets and for the private resources which we can investigate.
@nlsrchtr commented on GitHub (Mar 26, 2026):
It looks way better, I can create sites now!
But I observe a strange behaviour:
I can observe this only with newly created sites, existing sites are up and running. That's why I suspect, this might have to do with our changes.
@oschwartz10612 commented on GitHub (Mar 27, 2026):
This likely is because gerbil is using the old subnet on the interface.
Try to restart gerbil again now.
@nlsrchtr commented on GitHub (Mar 29, 2026):
Hi @oschwartz10612, you are right! After rebooting the gerbil service, everything was up and running!
Thanks a lot for your support, your patience and this wonderful piece of software!
Please keep up the good work on it!