CLUSTER-TOP

This is currently highly experimental. In the meantime checkout cluster-smi for GPUs,

The same as top but for multiple machines at the same time.

This use low-level C-code for efficiency for most operations. It only supports Linux (tested under Ubuntu)!

Output should be something like

+---------+---------------------------+-------+---------+-----------------+----------+
| Node    | RAM-Usage                 | Pid   | User    | Command         | CPU-Util |
+---------+---------------------------+-------+---------+-----------------+----------+
| node00  | 12349MiB / 32075MiB (37%) | 27133 | patwie  | sysbench        | 200%     |
|         |                           | 2546  | patwie  | audacity        | 8%       |
|         |                           | 3234  | patwie  | chrome          | 6%       |
|         |                           | 29589 | patwie  | plugin_host     | 4%       |
|         |                           | 30235 | patwie  | zsh             | 0%       |
|         |                           | 6147  | patwie  | chrome          | 0%       |
|         |                           | 25335 | patwie  | scdaemon        | 0%       |
|         |                           | 319   | patwie  | cluster-top-nod | 0%       |
|         |                           | 4873  | patwie  | chrome          | 0%       |
|         |                           | 29575 | patwie  | sublime_text    | 0%       |
+---------+---------------------------+-------+---------+-----------------+----------+

You might be interested as well in cluster-smi for GPUs.

Additional information are available, when using

user@host $ cluster-top -h

Usage of cluster-top:
  -t  show time of events

Monitoring Modes

This repository contains two versions: cluster-top-local, cluster-top.

Local (cluster-top-local)

cluster-top-local is the same as top:

Cluster (cluster-top)

cluster-top displays all information from cluster-top-local but for multiple machines at the same time.

On each machine you want to monitor you need to start cluster-top-node. They are sending information from the machine to a cluster-top-router, which further distributes these information to client (cluster-top) when requested.

You might be interested as well in cluster-top for CPUS.

Installation

Requirements + Dependencies

• ZMQ (4.0.1)

Unfortunately, ZMQ can only be dynamically linked (libzmq.so) to this repository and you need to build it separately by

# compile ZMQ library for c++
cd /path/to/your_lib_folder
wget http:/files.patwie.com/mirror/zeromq-4.1.0-rc1.tar.gz
tar -xf zeromq-4.1.0-rc1.tar.gz
cd zeromq-4.1.0
./autogen.sh
./configure
./configure --prefix=/path/to/your_lib_folder/zeromq-4.1.0/dist
make
make install

Finally:

export PKG_CONFIG_PATH=/path/to/your_lib_folder/zeromq-4.1.0/dist/lib/pkgconfig/:$PKG_CONFIG_PATH

Edit the CFLAGS, LDFLAGS in file nvvml/nvml.go to match your setup.

Compiling

You need to copy one config-file

user@host $ cp config.example.go config.go

To obtain a portable small binary, I suggest to directly embed the configuration settings (ports, ip-addr) into the binary as compile-time constants. This way, the app is fully self-contained (excl. libzmq.so) and does not require any configuration-files. This can be done by editing config.go:

...
c.RouterIp = "127.0.0.1"
c.Tick = 3
c.Timeout = 180
c.Ports.Nodes = "9080"
c.Ports.Clients = "9081"
...

Otherwise, you can specify the environment variable CLUSTER_TOP_CONFIG_PATH pointing to a yaml file (example in cluster-top.example.yml).

Then run

cd proc
go install
cd ..
make all

Run

1. start cluster-top-node at different machines
2. start cluster-top-router at a specific machine (machine with ip-addr: cluster_smi_router_ip)
3. use cluster-top like nvidia-top

Make sure, the machines can communicate using the specifiec ports (e.g., ufw allow 9080, 9081)

Use systemd

To ease the use of this app, I suggest to add the cluster-top-node into a systemd-service. An example config file can be found here. The steps would be

# add new entry to systemd
sudo cp docs/cluster-top-node.example.service /etc/systemd/system/cluster-top-node.service
# edit the path to cluster-top-node
sudo nano /etc/systemd/system/cluster-top-node.service
# make sure you can start and stop the service (have a look at you cluster-top client)
sudo service cluster-top-node start
sudo service cluster-top-node stop
# register cluster-top-node to start on reboots
sudo systemctl enable cluster-top-node.service

# last, start the service
sudo service cluster-top-node start