运行 GCP 虚拟机上的 MLFlow

Question

我已经在 GCP VM 实例上安装了 mlflow，现在我想用外部 IP 访问 mlflow UI。我尝试设置防火墙规则并打开 mlflow 的默认端口，但无法访问它。有人可以为 VM 实例上的运行 mlflow 提供分步过程吗？

Answer 1

我决定检查我的测试 VM 和运行 GCE VM 上的 mlflow 服务器。看看我下面的步骤：

基于 Ubuntu Linux 18.04 LTS

$ sudo apt update
$ sudo apt upgrade
$ cd ~
$ git clone https://github.com/mlflow/mlflow
$ cd mlflow
$ sudo apt install python3-pip
$ pip3 install mlflow
$ python3 setup.py build
$ sudo python3 setup.py install
$ mlflow --version
mlflow, version 1.7.1.dev0

运行 VM 实例内部 IP 上的 mlflow 服务器（默认 127.0.0.1）：

$ ifconfig 
ens4: flags=4163<UP,BROADCAST,RUNNING,MULTICAST>  mtu 1460
inet 10.XXX.15.XXX  netmask 255.255.255.255  broadcast 0.0.0.0
...

$ mlflow server --host 10.XXX.15.XXX
[2020-03-09 15:05:50 +0000] [8631] [INFO] Starting gunicorn 20.0.4
[2020-03-09 15:05:50 +0000] [8631] [INFO] Listening at: http://10.128.15.211:5000 (8631)
[2020-03-09 15:05:50 +0000] [8631] [INFO] Using worker: sync
[2020-03-09 15:05:50 +0000] [8634] [INFO] Booting worker with pid: 8634
[2020-03-09 15:05:51 +0000] [8635] [INFO] Booting worker with pid: 8635
[2020-03-09 15:05:51 +0000] [8636] [INFO] Booting worker with pid: 8636
[2020-03-09 15:05:51 +0000] [8638] [INFO] Booting worker with pid: 8638

从 VM 实例（从第二个连接）检查：

$ curl -I http://10.XXX.15.XXX:5000
HTTP/1.1 200 OK
Server: gunicorn/20.0.4
Date: Mon, 09 Mar 2020 15:06:08 GMT
Connection: close
Content-Length: 853
Content-Type: text/html; charset=utf-8
Last-Modified: Mon, 09 Mar 2020 14:57:11 GMT
Cache-Control: public, max-age=43200
Expires: Tue, 10 Mar 2020 03:06:08 GMT
ETag: "1583765831.3202355-853-3764264575"

set network tag mlflow-server

create firewall rule 允许访问端口 5000

$ gcloud compute --project=test-prj firewall-rules create mlflow-server --direction=INGRESS --priority=999 --network=default --action=ALLOW --rules=tcp:5000 --source-ranges=0.0.0.0/0 --target-tags=mlflow-server

从本地 Linux 计算机 nmap -Pn 35.225.XXX.XXX

检查

Starting Nmap 7.80 ( https://nmap.org ) at 2020-03-09 16:20 CET
Nmap scan report for 74.123.225.35.bc.googleusercontent.com (35.225.XXX.XXX)
Host is up (0.20s latency).
Not shown: 993 filtered ports
PORT     STATE  SERVICE
...
5000/tcp open   upnp
...

转到网络浏览器http://35.225.XXX.XXX:5000/

运行 GCP 虚拟机上的 MLFlow

Running MLFlow on GCP VM

python

google-cloud-platform

mlflow