#community-help

Server RAM Issue with Typesense Server

TLDR Bill raised concern about increasing RAM usage on running tests on Typesense. Jason suggested to upgrade to the version 0.23.0.rc69, which seemed to resolve the issue and increase the performance.

Powered by Struct AI

2

1

May 24, 2022 (19 months ago)
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:09 PM
Hello, I run a benchmark 10 concurrent reqs/s, total reqs 500 and the RAM of the server never shrinks. Every test adds up ~ 150mb and after 4 tests, the ram has stuck at 680 mb. Any idea?
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
12:31 PM
Are these requests writes or searches?
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
12:31 PM
searches
12:37
Bill
12:37 PM
● typesense-server.service - Typesense Server
Loaded: loaded (/etc/systemd/system/typesense-server.service; enabled; vendor>
Active: active (running) since Tue 2022-05-24 11:45:23 UTC; 43min ago
Docs: https://typesense.org
Main PID: 655 (typesense-serve)
Tasks: 61 (limit: 1131)
Memory: 691.6M
CGroup: /system.slice/typesense-server.service
└─655 /usr/bin/typesense-server --config=/etc/typesense/typesense-ser>
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
01:01 PM
This sounds very strange… May I know what version of Typesense this is on? Also, could you try this on 0.23.0.rc69?
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
01:02 PM
how can I check the version?
01:04
Bill
01:04 PM
I use version 0.23.0.rc6
01:04
Bill
01:04 PM
Kishore had created a deb package for me
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
01:05 PM
It's possible that has a memory leak.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
01:05 PM
should I install the v.0.23.0.rc69

1

01:05
Bill
01:05 PM
?
01:05
Bill
01:05 PM
is it available as deb package?
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
01:05 PM
In the link you have, if you change all the version numbers to rc69, you can get the deb for that
01:06
Jason
01:06 PM
Once you install this deb package, you want to make sure you restart the Typesense process
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
01:07 PM
Ok
01:10
Bill
01:10 PM
Ok I upgraded the node
01:10
Bill
01:10 PM
I'll perform the load test
01:12
Bill
01:12 PM
Jason Is this version faster than the previous?
01:12
Bill
01:12 PM
I perform 50 concurrent reqs/s in 1 GB RAM - 1 CPU and the latency is 300 ms on average
Jason
Photo of md5-8813087cccc512313602b6d9f9ece19f
Jason
01:37 PM
Yeah we've improved search performance in this version

1

01:38
Jason
01:38 PM
I'm surprised it works in 1 vCPU. In the past I've seen things stall with 1vcpu. You'd need at least 2 vCPU.
Bill
Photo of md5-be53735a2b0297bb542711c1d2ecea45
Bill
01:38 PM
That's perfect! I tested even with 120 concurrent requests on 1 GB RAM - 1 vCPU Shared droplet and the latency is ~ 500ms
01:39
Bill
01:39 PM
The ram consumption is 30-32mb now with 120 conc reqs per sec

1

Typesense

Lightning-fast, open source search engine for everyone | Knowledge Base powered by Struct.AI

Indexed 3005 threads (79% resolved)

Join Our Community

Similar Threads

Memory Leak Issue in Typesense Server v0.21.0

Jim was experiencing a potential memory leak while load testing an API. Jason and Kishore Nallan suggest it may be a known issue in v0.22.0 RC builds which is resolved in later versions. Jim identifies a specific search filter escalating the issue. Kishore Nallan confirms a fix in the 0.22 RC builds.

4

30
27mo

Discussion on Performance and Scalability for Multiple Term Search

Bill asks the best way for multi-term searches in a recommendation system they developed. Kishore Nallan suggested using embeddings and remote embedder or storing and averaging vectors. Despite testing several suggested solutions, Bill continued to face performance issues, leading to unresolved discussions about scalability and recommendation system performance.

3

105
1w

Discussing RAM, CPU, and Hosting Options for Project

Bill asked about plans, domains and performance disparity with Typesense versus Digital Ocean. Kishore Nallan and Jason provided insights about network and disk performances, comparisons and prices.

1

30
11mo
Solved

Optimizing Typesense Implementation for Large Collections

Oskar faced performance issues with his document collection in Typesense due to filter additions. Jason suggested trying a newer Typesense build and potentially partitioning the data into country-wise collections. They also discussed reducing network latency with CDN solutions.

5

67
11mo
Solved

Typesense Server Crash and Counting Issue

Dan reported a server crash and counting issues with Typesense server version 0.26.0.rc1. Jason suggested an upgrade to `0.26.0.rc13` while Kishore Nallan committed to investigate the issue based on a sample dataset from Dan.

11
3mo