Updates to onprem docs (#19275)
Mateusz Bularz committed
Dec 23, 2024 at 15:34 UTC
5897fe083853e2969ef82aba1571b90244f17146
2 files changed
+64
-12
docs/netdata-cloud/netdata-cloud-on-prem/poc-without-k8s.md
+17
-11
@@ -28,19 +28,21 @@ sudo ./provision.sh install \
28
The script above is responsible for:
29
30
1. Prompting the user to provide:
31
- - `-key-id` - AWS ECR access key ID.
32
- - `-access-key` - AWS ECR Access Key.
33
- - `-onprem-license-key` - Netdata Cloud On-Prem license key.
34
- - `-onprem-license-subject` - Netdata Cloud On-Prem license subject.
35
- - `-onprem-url` - URL for the On-prem (without http(s) protocol).
36
- - `-certificate-path` - path to your PEM encoded certificate.
37
- - `-private-key-path` - path to your PEM encoded key.
31
+
32
+ - `-key-id` - AWS ECR access key ID.
33
+ - `-access-key` - AWS ECR Access Key.
34
+ - `-onprem-license-key` - Netdata Cloud On-Prem license key.
35
+ - `-onprem-license-subject` - Netdata Cloud On-Prem license subject.
36
+ - `-onprem-url` - URL for the On-prem (without http(s) protocol).
37
+ - `-certificate-path` - path to your PEM encoded certificate.
38
+ - `-private-key-path` - path to your PEM encoded key.
39
40
2. Installation will begin. The script will install:
40
- - Helm
41
- - Kubectl
42
- - AWS CLI
43
- - K3s cluster (single node)
41
+
42
+ - Helm
43
+ - Kubectl
44
+ - AWS CLI
45
+ - K3s cluster (single node)
46
47
3. The script starts to provision the K3s cluster with gathered data.
48
@@ -67,3 +69,7 @@ To uninstall the whole PoC, use the same script that installed it, with the `uni
69
cd <script dir>
70
sudo ./provision.sh uninstall
71
```
72
+
73
+> **Note**
74
+>
75
+> In most cases of a failed installation, an automatic prompt will trigger for the user to start the uninstallation process automatically.
docs/netdata-cloud/netdata-cloud-on-prem/troubleshooting.md
+47
-1
@@ -11,12 +11,58 @@ These components should be monitored and managed according to your organization'
11
12
## Common Issues
13
14
+### Installation cannot finish
15
+
16
+If you are getting error like:
17
+
18
+```
19
+Installing netdata-cloud-onprem (or netdata-cloud-dependency) helm chart...
20
+[...]
21
+Error: client rate limiter Wait returned an error: Context deadline exceeded.
22
+```
23
+
24
+There are probably not enough resources available. Fortunately, it is very easy to verify with the `kubectl` utility. In the case of a full installation, switch the context to the cluster where On-Prem is being installed. For the Light PoC installation, SSH into the Ubuntu VM where `kubectl` is already installed and configured.
25
+
26
+To verify check if there are any `Pending` pods:
27
+
28
+```shell
29
+kubectl get pods -n netdata-cloud | grep -v Running
30
+```
31
+
32
+To check which resource is a limiting factor pick one of the `Pending` pods and issue command:
33
+
34
+```shell
35
+kubectl describe pod <POD_NAME> -n netdata-cloud
36
+```
37
+
38
+At the end in an `Events` section information about insufficient `CPU` or `Memory` on available nodes should appear.
39
+Please check the minimum requirements for your on-prem installation type or contact our support - `support@netdata.cloud`.
40
+
41
+> **Warning**
42
+>
43
+> In case of the Light PoC installations always uninstall before the next attempt.
44
+
45
+### Installation finished but login does not work
46
+
47
+It depends on the installation and login type, but the underlying problem is usually located in the `values.yaml` file. In the case of Light PoC installations, this is also true, but the installation script fills in the data for the user. We can split the problem into two variants:
48
+
49
+1. SSO is not working - you need to check your tokens and callback URLs for a given provider. Equally important is the certificate - it needs to be trusted, and also hostname(s) under `global.public` section - make sure that FQDN is correct.
50
+2. Mail login is not working:
51
+ 1. If you are using a Light PoC installation with MailCatcher, the problem usually appears if the wrong hostname was used during the installation. It needs to be a FQDN that matches the provided certificate. The usual error in such a case points to a invalid token.
52
+ 2. If the magic link is not arriving for MailCatcher, it's likely because the default values were changed. In the case of using your own mail server, check the `values.yaml` file in the `global.mail.smtp` section and your network settings.
53
+
54
+If you are getting the error `Something went wrong - invalid token` and you are sure that it is not related to the hostname or the mail configuration as described above, it might be related to a dirty state of Netdata secrets. During the installation, a secret called `netdata-cloud-common` is created. By default, this secret should not be deleted by Helm and is created only if it does not exist. It stores a few strings that are mandatory for Netdata Cloud On-Prem's provisioning and continuous operation. Because they are used to hash the data in the PostgreSQL database, a mismatch will cause data corruption where the old data is not readable and the new data is hashed with the wrong string. Either a new installation is needed, or contact to our support to individually analyze the complexity of the problem.
55
+
56
+> **Warning**
57
+>
58
+> If you are changing the installation namespace secret netdata-cloud-common will be created again. Make sure to transfer it beforehand or wipe postgres before new installation.
59
+
60
### Slow Chart Loading or Chart Errors
61
62
When charts take a long time to load or fail with errors, the issue typically stems from data collection challenges. The `charts` service must gather data from multiple Agents within a Room, requiring successful responses from all queried Agents.
63
64
| Issue | Symptoms | Cause | Solution |
19
-|----------------------|-----------------------------------------------------------------------------------------------------------------|------------------------------------------------------------------------------|-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|
65
+| -------------------- | --------------------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
66
| Agent Connectivity | - Queries stall or timeout<br/>- Inconsistent chart loading | Slow Agents or unreliable network connections prevent timely data collection | Deploy additional [Parent](/docs/observability-centralization-points/README.md) nodes to provide reliable backends. The system will automatically prefer these for queries when available |
67
| Kubernetes Resources | - Service throttling<br/>- Slow data processing<br/>- Delayed dashboard updates | Resource saturation at the node level or restrictive container limits | Review and adjust container resource limits and node capacity as needed |
68
| Database Performance | - Slow query responses<br/>- Increased latency across services | PostgreSQL performance bottlenecks | Monitor and optimize database resource utilization:<br/>- CPU usage<br/>- Memory allocation<br/>- Disk I/O performance |