zfs.rent

← All updates

Customer IP Address Reassignment

tl;dr

Customer IPv4 addresses will be renumbered to fit inside a single /24 prefix.

The migration is scheduled for 2025-11-24 (17:00 - 23:00 PST).

We will be BGP originating the new prefix, improving network reliability and speed.

  1. Added transit redundancy. Customer VMs remain available during single-carrier faults.
  2. Increased bandwidth to select providers. Our peering links are 10 Gbps.
  3. Future outage mitigation. We can solve more issues remotely.

Existing customers should have already received the maintenance window email.

On 2025-11-14, we will migrate an initial batch. Email [email protected] to opt-in.


More background:

After the 2025-10-13 outage, we moved our network redundancy plan to high priority. The last two outages were caused by a faulty Ethernet crimp in the datacenter's patch panel (not on our side!) Basically, they kinda effed up the uplink cable during our August cabinet move. The link quality was marginal — during Ethernet auto-negotiation, it had a 50/50 chance of actually coming online (but we didn't know that). This fluke of a cable caught us off guard twice.1

Relying on a single transit provider made it virtually impossible to resolve the cabling issue remotely. We had no idea what the issue was — and the on-site technicians couldn't try much other than power-cycling.2

On the bright side, it motivated us to finalize our private interconnects and AS peering.


The new connectivity will help us resolve network issues quicker.

More importantly, customer instances will be accessible during single-carrier outages
(such as the two mentioned above).

And last but not least, some destinations will have 10G paths!


We will be originating the new customer prefixes3 from AS38439.

Our BGP sessions have been online for a month, and we have a few customers already using the subnet. So far, we are connected to two transit providers and two Internet Exchanges.

View our current peers here: bgp.he.net/AS38439.

Also, random side note... we now have IPv6 reachability to Cogent ISP 🙂4


As always, please reach out to [email protected] with any questions or comments.

Scott and I are always happy to entertain feedback / solve issues.

Last point:

I am generally not a fan of bulk IP renumbering customers —
and wouldn't make this move unless it was absolutely best for the service.


Keeping your bits safe since 2020!

Started in Sacramento, ended up in Fremont, and now we have an ASN!

-- Ryan


P.S. In other news, we also finally have functional SMART support working. Enabling support is still a manual process, so it's only active for less than 10 customers. We hope to roll it out globally by end-of-year.

We still have more update posts planned. We will create a mailing list for them, since they aren't related to customer VM maintenance.


  1. Fool me twice, shame on me :/ 

  2. In hindsight, I should have asked them to verify the uplink cable. But it didn't come to mind— because the new cabinet's network had already been stable for a month. After I drove down, my last-ditch troubleshooting idea was to plug my laptop into the transit drop— and voilà bad cable. 

  3. 199.127.129.0/24 and 2602:F831::/48 

  4. Cogent and Hurricane Electric have a long-standing dispute and do not exchange IPv6 prefixes. See bgp.tools/kb/partitions. With Ryamer transit, we can finally reach Cogent IPv6.