NODEBB: Nginx error performance & High CPU

[phenomlab]

@DownPW said in NODEBB: Nginx error performance & High CPU:

everything is in the nginx.conf and the config.json if I understand correctly

Yes, that’s correct

@DownPW said in NODEBB: Nginx error performance & High CPU:

In terms of disk space, memory, cpu what does it give ?

There should be little change in terms of the usage. What clustering does is essentially provide multiple processes to carry out the same tasks, but obviously much faster than one process only ever could.

@DownPW said in NODEBB: Nginx error performance & High CPU:

After all of that, we need to restart nodebb I imagine ?

Correct

@DownPW said in NODEBB: Nginx error performance & High CPU:

These new ports (4567", “4568”, "4569) must also be open in the Hetzner interface?

Will not be necessary as they are not available publicly, but only to the reverse proxy on 127.0.0.1

[DownPW]

OK.

@phenomlab

I resume in details.

1- Stop nodebb
2- Stop iframely
3- Stop nginx
4- Install redis server : sudo apt install redis-server
5- Change nodebb Config.json file (can you verifiy this synthax please ? nodebb documentation tell “database”: 0 and not “database”: 5 - but maybe it’s just a name and i can use the same as mongo like “database”: nodebb , I moved the port directive) :

{
    "url": "https://XX-XX.net",
    "secret": "XXXXXXXXXXXXXXXX",
    "database": "mongo",
   "port": [4567, 4568,4569],
    "mongo": {
        "host": "127.0.0.1",
        "port": "27017",
        "username": "XXXXXXXXXXX",
        "password": "XXXXXXXXXXX",
        "database": "nodebb",
        "uri": ""
    },
    "redis": {
        "host":"127.0.0.1",
        "port":"6379",
        "database": 5
    }
}

6- Change nginx.conf :

# add the below block for nodeBB clustering
upstream io_nodes {
    ip_hash;
    server 127.0.0.1:4567;
    server 127.0.0.1:4568;
    server 127.0.0.1:4569;
}

server {
	server_name XX-XX.net www.XX-XX.net;
	listen XX.XX.XX.X;
	listen [XX:XX:XX:XX::];
	root /home/XX-XX/nodebb;
	index index.php index.htm index.html;
	access_log /var/log/virtualmin/XX-XX.net_access_log;
	error_log /var/log/virtualmin/XX-XX.net_error_log;

# add the below block which will force all traffic into the cluster when referenced with @nodebb
    
location @nodebb {
     proxy_pass http://io_nodes;
    }

        location / {
				
		limit_req zone=flood burst=100 nodelay; 
		limit_conn ddos 10; 
		proxy_read_timeout 180; 

                proxy_set_header X-Real-IP $remote_addr;
                proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
                proxy_set_header Host $http_host;
                proxy_set_header X-NginX-Proxy true;
               # It's necessary to set @nodebb here so that the clustering works
                proxy_pass @nodebb;
                proxy_redirect off;

                # Socket.IO Support
                proxy_http_version 1.1;
                proxy_set_header Upgrade $http_upgrade;
                proxy_set_header Connection "upgrade";
        }

	listen XX.XX.XX.XX:XXssl http2;
	listen [XX:XX:XX:XX::]:443 ssl http2;
	ssl_certificate /etc/ssl/virtualmin/166195366750642/ssl.cert;
	ssl_certificate_key /etc/ssl/virtualmin/166195366750642/ssl.key;
	if ($scheme = http) {
		rewrite ^/(?!.well-known)(.*) "https://XX-XX.net/$1" break;
	}
}

7- restart redis server systemctl restart redis-server.service
8- Restart nginx
9- Restart iframely
10- Restart nodebb
11- test configuration

[phenomlab]

@DownPW said in NODEBB: Nginx error performance & High CPU:

5- Change nodebb Config.json file (can you verifiy this synthax please ? nodebb documentation tell “database”: 0 and not “database”: 5,

All fine from my perspective - no real need to stop iFramely, but ok. The database number doesn’t matter as long as it’s not being used. You can use 0 if you wish - it’s in use on my side, hence 5.

[phenomlab]

@DownPW I’d add another set of steps here. When you move the sessions away from mongo to redis you are going to encounter issues logging in as the session tables will no longer match meaning none of your users will be able to login

To resolve this issue

Review https://sudonix.com/topic/249/invalid-csrf-on-dev-install and implement all steps - note that you will also need the below string when connecting to mongodb

mongo -u admin -p <password> --authenticationDatabase=admin

Obviously, substitute <password> with the actual value. So in summary:

1. Open the mondogb console
2. Select your database - in my case use sudonix;
3. Issue this command db.objects.update({_key: "config"}, {$set: {cookieDomain: ""}});
4. Press enter and then type quit() into the mongodb shell
5. Restart NodeBB
6. Clear cache on browser
7. Try connection again

[DownPW]

@phenomlab

Hmm ok when perform these steps ?

[phenomlab]

@DownPW After you’ve setup the cluster and restarted NodeBB

[crazycells]

@phenomlab said in NODEBB: Nginx error performance & High CPU:

@crazycells said in NODEBB: Nginx error performance & High CPU:

4567, 4568, and 4569… Is your NodeBB set up this way?

It’s not (I set their server up). Sudonix is not configured this way either, but from memory, this also requires redis to handle the session data. I may configure this site to do exactly that.

yes, you might be right about the necessity. We have redis installed.

[crazycells]

@DownPW yes, this is the place to start:

https://docs.nodebb.org/configuring/scaling/

[crazycells]

@DownPW since you pointed it out, I just remembered. Since we know when this crowd will come and be online on our forum, that particular day, we switch off iframely and all preview plugins. That also helps to open the pages faster.

[crazycells]

@phenomlab a general but related question. Since opening three ports help, is it possible to increase this number? For example, can we run 5 ports NodeBB at the same time to smooth the web page experience; or is “3” goldilocks number for maximum efficiency?

[phenomlab]

@crazycells It’s not necessarily the “Goldilocks” standard - it really depends on the system resources you have available. You could easily extend it as long as you allow for the additional port(s) in the nginx.conf file also.

Personally, I don’t see the need for more than 3 though.

[DownPW]

@phenomlab

Ok redis is ok now. Thanks for your help

I would like know to obtain the connecting clients Real IP on Nginx log.

I read I have need ngx_http_realip_module for nginx but not active by default but I don’t know if virtualmin have this module enabled.

[DownPW]

@DownPW said in NODEBB: Nginx error performance & High CPU:

@phenomlab

Ok redis is ok now. Thanks for your help

I would like know to obtain the connecting clients Real IP on Nginx log.

I read I have need ngx_http_realip_modulefor nginx but not active by default but I don’t know if virtualmin have this module enabled.

EDIT: OK it will be enabled by default on virtualmin :

nginx -v

[DownPW]

@phenomlab

I have activate ngx_http_realip_module on /etc/nginx/nginx.conf on http block like this :

It seems to be good for you ?

[phenomlab]

@DownPW yes, that looks fine.

[DownPW]

@phenomlab

We have sometimes this error, what do you think about it ?

[phenomlab]

@DownPW I suspect that’s a failure of the socket server to talk to redis, but the NodeBB Devs would need to confirm.

[DownPW]

@phenomlab

Seems to be better with some scaling fix for redis on redis.conf. I haven’t seen the message yet since the changes I made

# I increase it to the value of /proc/sys/net/core/somaxconn
tcp-backlog 4096 


# I'm uncommenting because it can slow down Redis. Uncommented by default !!!!!!!!!!!!!!!!!!! 
#save 900 1
#save 300 10
#save 60 10000

If you have other Redis optimizations. I take all your advice

https://severalnines.com/blog/performance-tuning-redis/