# Support of ip range as primary key
**URL:** https://forum.scylladb.com/t/support-of-ip-range-as-primary-key/1927
**Category:** ScyllaDB
**Tags:** data-model, cassandra, vector-search
**Created:** [May 1, 2024, 10:28am UTC](https://forum.scylladb.com/t/support-of-ip-range-as-primary-key/1927 "2024-05-01T10:28:03Z")
**Posts on this page:** 4
**Page:** 1
### Author:  [@PyramidPlayer](https://forum.scylladb.com/u/PyramidPlayer)
#### Post date: [May 1, 2024, 10:28am UTC](https://forum.scylladb.com/t/support-of-ip-range-as-primary-key/1927/1 "2024-05-01T10:28:03Z")
Hi, Community.
I’m currently working on storing a table with IP ranges as the partition key (`ip_from` and `ip_to` declared as inet type). Since IP ranges may overlap, I’ve added an additional unique clustering key `row_num`, to ensure that I can always select the same row from a set of ranges. The create table statement looks like this:
create table my\_table ( ip\_from inet, ip\_to inet, row\_num bigint, …, primary key ((ip\_from, ip\_to), row\_num) )
I’m executing queries to find a row with an IP range that includes a provided IP address:
select \* from my\_table where ip\_from \<= ? and ip\_to \>= ? limit 1 allow filtering
In some cases, I manage to find the row I’m looking for (for example, any IP in the range `2607:fb90:e200:8000::` to `2607:fb90:e200:80ff:ffff:ffff:ffff:ffff`), but in other cases, nothing is found (for example, any IP in the range `2603:2000::` to `2603:203f:ffff:ffff:ffff:ffff:ffff:ffff`).
Could anyone explain this behavior?
(I understand that this might not be the most appropriate task to solve using Scylla, but perhaps there’s a way to tackle it somehow)
UPD: I found workaround hack but I don’t think that it is best solution. I converted all ip ranges to CIDR format and used 2 parts of it as 2 primary keys. For example, ip range `2607:fb90:e200:8000::` - `2607:fb90:e200:80ff:ffff:ffff:ffff:ffff` can be presented as `2607:fb90:e200:8000::/56` - `2607:fb90:e200:8000::` as first key and `56` as second key. It is good, if all ranges has the same ip prefix in CIDR, but in my case this is not the true. I have to iterate with passed ip to find needed ip\_prefix from 128 to 1 for IPv6 and from 32 to 1 for IPv4.
It looks like the best solution is to use vector search mechanism like Cassandra supports now, but as I know Scylla don’t support it yet.
---
### Author:  [@avikivity](https://forum.scylladb.com/u/avikivity)
#### Post date: [July 9, 2024, 1:08pm UTC](https://forum.scylladb.com/t/support-of-ip-range-as-primary-key/1927/2 "2024-07-09T13:08:23Z")
Can you post an example that fails?
---
### Author:  [@PyramidPlayer](https://forum.scylladb.com/u/PyramidPlayer)
#### Post date: [July 9, 2024, 1:33pm UTC](https://forum.scylladb.com/t/support-of-ip-range-as-primary-key/1927/3 "2024-07-09T13:33:20Z")
There is no fail or error. Just nothing found while I know that the record exists. If I execute query like:
```auto
select * from my_table where ip_from = ? and ip_to = ?
```
and pass `2607:fb90:e200:8000::` for ip\_from and `2607:fb90:e200:80ff:ffff:ffff:ffff:ffff` for ip\_to then I will find the record with this range. But any other query that is looking for range by specific ip, that this range includes, like:
```auto
select * from my_table where ip_from <= ? and ip_to >= ? limit 1 allow filtering
```
will return nothing as soon as the table size becomes more than 100 000 records.
---
### Author:  [@avikivity](https://forum.scylladb.com/u/avikivity)
#### Post date: [October 6, 2024, 10:53am UTC](https://forum.scylladb.com/t/support-of-ip-range-as-primary-key/1927/4 "2024-10-06T10:53:34Z")
Please post a script that creates such a table, populates it, and runs the failing query. Best in a new issue.