Monitoring And Log Management Teach You How To Maintain The Health Of Alibaba Cloud Servers In Singapore

1.

Confirm the area and resource list (preparation).

Subsection 1: Log in to the Alibaba Cloud console and select the "Singapore (ap-southeast-1)" area in the upper right corner.
Subsection 2: List the ECS instance IDs, intranet IPs, key services running (nginx, mysql, etc.), and log paths (/var/log/nginx/*.log).
Subsection 3: Prepare the SSH private key for the account with console permissions (cloud monitoring and log service permissions) and the instance.

2.

Enable cloud monitoring and create monitoring items

Subsection 1: Console path: Cloud Monitoring > Host Monitoring, confirm that the ECS instance has been recognized by Cloud Monitoring (if not recognized, go to "Install Monitoring Plugin").
Subsection 2: Select the required basic metrics (CPUUtilization, MemoryUtilization, DiskUsage, NetIn/NetOut) and enable them for each instance. If there is no memory metric by default, you need to install a monitoring plugin to collect memory.

3.

Install the monitoring and log collection agent (practical steps).

Section 1: Cloud Monitoring Plugin: Go to the cloud monitoring console's host monitoring page, click "Install Monitoring Plugin," select Linux/Windows, follow the console prompts to execute the one-click installation script curl or wget, and run it.
Subsection 2: Log Service: Console > Log Service > Create a Project and Logstore, then click "Install Logtail" in Logstore, download the Logtail package for the corresponding operating system or copy the one-click installation command provided by the console (including Project/Logstore token), and execute it on ECS to complete registration.

4.

Configure Logtail collection tasks (specific example).

Segment 1: In the Logtail console, click "Configure Acquisition Path," enter the local log path, such as /var/log/nginx/access.log, and set parsing rules (commonly in nginx format or JSON).
Subsection 2: Set collection tags (instanceId, app, env) and test the collection sample in the console to confirm logs can be retrieved from Logstore.

5.

Create alert strategies and notifications (thresholds and durations).

Subsection 1: Cloud Monitoring > Alerts > Creating Alert Rules: Select metrics (e.g., CPU % > 80% lasting 5 minutes), select monitoring objects (by instance ID or label).
Subsection 2: Set notification methods: Email, SMS, Alibaba Cloud SMS/DingTalk Webhook/WeCom, or custom callback URL. It is recommended to configure contact groups and test notifications.

6.

Automated response plan (recommended implementation steps).

Subsection 1: When triggering an alarm, it is recommended to first call Function Compute or Webhook. The Webhook backend can be a lightweight operations script to execute tasks such as restarting the service (systemctl restart nginx) or collecting diagnostic snapshots.
Subsection 2: Implementation method: In the alarm action, select "Call HTTP Callback"; the callback triggers a controlled intermediate service (authentication required), which executes recovery commands on the target instance via SSH or Cloud Assistant.

7.

On-site inspection of commonly used commands and log locations

Subsection 1: Real-time check: top/htop, free -m, df -h, iostat -xm 5 3, ss -tunlp, journalctl -u service name.
Subsegment 2: Log pull: tail -n 200 /var/log/nginx/error.log; If you need to query in Log Service, use Log Search to retrieve and download the exception segment by time and keyword.

8.

Q: How can I quickly confirm whether surveillance is effective in Singapore?

Q: How do I confirm that monitoring is effective? Answer: Log in to Cloud Monitoring, select the Singapore region, open the instance monitoring page, and confirm that the CPU/memory/disk charts have real-time data; Check in Log Service whether the corresponding Logstore has the latest log entries; Loads can be manually created (stress commands) or test logs written to verify alarms and log collection.

9.

Q: How can false alarms be reduced? What practical strategies are there?

Q: How can false positives be reduced? Answer: Set reasonable thresholds and continuous cycles (for example, trigger an alarm only after three consecutive sampling exceeds the threshold), use tag grouping to precisely select instances, combine log conditions (only trigger when a specific error log occurs and CPU is high), and configure suppression windows and alarm grading in the alert policy.

10.

Q: How can you quickly recover and retain diagnostic data when a fault occurs?

Q: How can I recover quickly and keep my diagnosis? Answer: After triggering the alarm, an automated script is called via webhook to complete one-click collection (top/ps/df/iostat, core dump, and related logs packaged into OSS), and the service is attempted to automatically restart the service; At the same time, the diagnostic package is uploaded to OSS and logged in the SLS for easier subsequent auditing and issue traceability.