CGRateS does not preload filter indexes (attribute_filter_indexes / supplier_filter_indexes)

[Victor Shulha]

Hi All, 

We’d like to discuss an issue we are seeing with CGRateS on two versions:

• v0.10.4~dev-20240610165520-cc7210bbeee7

• v0.11.0~dev-20250922111205-5894e0d7f349

---

Environment:

• CGRateS + Kamailio node: AMD EPYC 7R13 (16 vCPU @ 3.6 GHz), 32 GB RAM, AWS gp3 SSD, AWS Nitro V5 (~12.5 Gbps)

• Redis 7.1 node: AWS Graviton2 (2 vCPU @ 2.5 GHz), 2 GB RAM, in-memory only, AWS Nitro (~5 Gbps)

---

Issue description

After a "cold" start, running LoadCache correctly loads static caches, but the following indexes remain empty:

• attribute_filter_indexes

• supplier_filter_indexes

Screenshot 1: shows how these indexes are enabled in the configuration, but stay at zero after LoadCache is completed.

These indexes start to grow only during live Session.Auth traffic.
As they grow (up to ~300k entries), CGRateS speeds up accordingly.

Screenshot 2: shows the dynamic growth of attribute_filter_indexes, supplier_filter_indexes

---

Impact

For the first ~10–15 minutes after CGRates restart, even with ~60 CPS, we get SIP authentication timeouts from Kamailio via CGR agent.

Example log:

{"Event":"PBX_SESSION_AUTH_REPLY","Error":"ATTRIBUTES_ERROR:context deadline exceeded"}

Once the indexes are fully populated, CGRateS handles 100+ CPS without issues.

---

Question

Is there any way to make CGRateS prebuild attribute_filter_indexes and supplier_filter_indexes at startup, so we don’t have the long warm-up period with SIP auth timeouts?

[Ionut Boangiu]

Hi Victor,

Could you please open an issue where you also provide the complete reproducing steps?

To answer your question: as you're doing now should work (only on v0.11), otherwise you can also set "precache" to true for the specific partitions that you want to be loaded on startup.

Thanks,
Ionuț

[Victor Shulha]

Hi Ionuț,

Thanks for the hints. We will try to test them and, in case of any difficulties, will create a report on GitHub.