NtopNG does not start when Ram Disks for tmp and var are on

jjstecchino

If I turn on Ram disks NtopNg does not start after a reboot. Fresh install of NtopNG starts ok, but doesnt start after reboot. Logs show:

Mar 22 08:13:13	php-fpm	51795	/rc.start_packages: The command '/usr/local/bin/redis-cli SET ntopng.user.admin.password 'f9f16d97c90d8c6f2cab37bb6d1f1992'' returned exit code '1', the output was 'Could not connect to Redis at 127.0.0.1:6379: Operation timed out Could not connect to Redis at 127.0.0.1:6379: Operation timed out'

jjstecchino

It would appear that /var/db/redis does not get saved/restored on reboot. without /var/db/redis redis fails to start and without it NtopNG cant run.

After manually creating that directory and starting the redis service NtopNG starts as expected but of course the /var/db/redis does not survive reboot.

Can this be fixed in the ramdisk save/restore script run at boot/reboot?

Thanks

pfBasic

As a workaround for now you could use the package shellcmd:

If it just needs the directory to be there on boot:

Then setup a shellcmd to```
mkdir /var/db/redis


If it needs the contents of the directory, create a directory to store the temp files:

mkdir /usr/local/etc/ntopng_tmp


then before you shutdown (or as a cronjob):

cp -iprv /var/db/redis /usr/local/etc/ntopng_tmp


and as a shellcmd:

cp -iprv /usr/local/etc/ntopng_tmp /var/db/redis


or if the directory also must be created each time:

mkdir /var/db/redis && cp -iprv /usr/local/etc/ntopng_tmp /var/db/redis

jjstecchino

I'll give shellcmd a try, hopefully it will try to create the directory after the ram disk is created but before redis seervice gets started.

I just wanted to make the developer aware of the bug so that it can be fixed.

I dont know if I can file a bug myself. Over the years I have always described problem I found here in the forum and the dev took care of it.

Thanks for taking time to answer

jjstecchino

Shellcmd or earlycmd does not work for me. Maybe wrong timing in the directory creation.

However manuallly editing /etc/pfSense-rc like this works well. I gues this is the fix that could be implemented. Of course it does not save/restore any data and perhaps there is a better place to create this dir

make some directories in /var

/bin/mkdir -p $varrunpath /var/log /var/etc /var/db/entropy /var/db/rrd /var/at/jobs/ /var/empty /var/db/ntopng /var/log/nginx
2>/dev/null
chmod 0555 /var/empty
/bin/rm -rf $varrunpath/*

doktornotor

This package will NOT work properly with the ramdisks. It is also a rather horrible idea if you ask me.

# du -sh /var/db/ntopng/
4.5G    /var/db/ntopng/

The above is data accumulated after some ~6 months of usage on a site with pretty moderate traffic (~100GB/month). Sure like hell you do not want that to be in your RAM, or be backing that up and restoring on reboot.

jjstecchino

@doktornotor:

This package will NOT work properly with the ramdisks. It is also a rather horrible idea if you ask me.
# du -sh /var/db/ntopng/
4.5G    /var/db/ntopng/
The above is data accumulated after some ~6 months of usage on a site with pretty moderate traffic (~100GB/month). Sure like hell you do not want that to be in your RAM, or be backing that up and restoring on reboot.

No I actually don't care about historical data but I find ntopng quite useful to have a glimpse on the traffic that is going through pfsense in a nicely presented interface. I am ok with the db being cleared on reboot. I dont know though what will happen if var runs out of space, i.e. will dhcpd be unable to write leases on disk etc.