Skip to content

Commit 6e114ab

Browse files
pmigcdxker
authored andcommitted
docs: improved the glasskube installation instructions
Signed-off-by: Philip Miglinci <pmig@glasskube.com>
1 parent f1d0d8c commit 6e114ab

File tree

1 file changed

+55
-3
lines changed

1 file changed

+55
-3
lines changed

glasskube/README.md

Lines changed: 55 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -4,9 +4,12 @@
44
## 1. Start a new minikube cluster
55

66
[Minikube](https://minikube.sigs.k8s.io/docs/start/) can be installed via Homebrew or binary downloads.
7+
Also make sure to enable the ingress and metrics plugin
78

89
```shell
910
minikube start -p trieve
11+
minikube addons enable metrics-server -p trieve
12+
minikube addons enable ingress -p trieve
1013
```
1114

1215
## 2. Bootstrap glasskube
@@ -68,15 +71,64 @@ kubectl port-forward svc/trieve-keycloak-service 8080:8080
6871

6972
Open [dashboard.localtrieve.com](http://dashboard.localtrieve.com) in your browser.
7073

71-
## Optional commands
74+
## Optional commands & debugging info
7275

73-
#### Scale down all embedding servers:
76+
#### Working on low CPU machines
77+
78+
You can scale down all embedding servers to reduce the memory and cpu usage of trieve.
79+
80+
```shell
81+
kubectl scale deployment trieve-embedding-bgem3 trieve-embedding-jina trieve-embedding-reranker trieve-embedding-splade-doc trieve-embedding-splade-query --replicas 0
82+
```
83+
84+
If you still want to do end-to-end test you can only start the `trieve-embedding-bgem3`
85+
and configure the dataset in a specific way:
86+
87+
1. Create a new dataset
88+
2. Disable full text search
89+
3. Import test data
90+
91+
After that the ingestion worker will orchestrate the ingestion.
92+
93+
You can see the ingestion progress by connecting to the redis pod and execute the `redis-cli` and
94+
monitor progress by looking at the queues `llen ingestion` and `llen processing`.
95+
96+
After ingestion is done you can search the dataset by only setting the search type to semantic so
97+
only bgem3 is used.
98+
99+
#### Qdrant
100+
101+
Qdrant throws some weird exceptions during the initial database migrations, but keeps working normally after.
102+
103+
You can monitor the status of the collections by looking at the Qdrant ui:
104+
105+
```shell
106+
glasskube open qdrant
107+
```
108+
109+
You can fetch the API key with:
110+
111+
```
112+
kubectl get secret trieve-qdrant-qdrant-apikey -o jsonpath='{.data.api-key}' | base64 -d
113+
```
114+
115+
#### ClickHouse
116+
117+
In order to test if ClickHouse works you can connect to clickhouse via:
74118

75119
```shell
76-
kubectl scale deployment trieve-embedding-bgem3 trieve-embedding-jina trieve-embedding-reranker trieve-embedding-splade-doc trieve-embedding-splade-query --replicas 0
120+
kubectl exec -it chi-trieve-clickhouse-cluster1-0-0-0 -- clickhouse-client
121+
```
122+
123+
The password is currently `password` but will be changed in the future.
77124

125+
In order to see if the custom embedding function works you can execute:
126+
127+
```
128+
select embed_p('test');
78129
```
79130

131+
80132
#### Deleting the cluster
81133

82134
```shell

0 commit comments

Comments
 (0)