Posts tagged with “sysadmin”

Young me doing IT stuff

Written by Simone

written notes about msdos

Obsolete major version 13 (uh oh!)

Written by Simone

A package configuration text screen by Debian apt which is informing the user that postgresql packages are old and need to be upgraded.

Yesterday after installing some new packages I was greeted by this kind reminder 😀

I began stopping services which use a Postgresql database and even forgot about Dendrite.. Nothing so bad as I imagined, tho, just run the suggested commands and everything got up and running in a few minutes.

Debian rocks! 😍

Notes

Written by Simone

I'll update this post whenever I see fits, A.K.A. when I have other bits of information which don't require a whole post.

Operations on files and directories

Move files and directories to the current/parent directory in Linux

Current dir

find . -type f -exec mv {} . \;

Parent dir

find . -type f -exec mv {} .. \;

Recursive chmod on files and dirs

To change all the directories to 755 (-rwxr-xr-x):

find /var/www/blog -type d -exec chmod 755 {} \;

To change all the files to 644 (-rw-r--r--):

find /var/www/blog -type f -exec chmod 644 {} \;

Certbot common commands

Register single domain:

certbot certonly --standalone -d domain.tld --dry-run

Renew single domain:

certbot renew --cert-name domain.tld --dry-run
  • Remove “–dry-run” when ok.

Revoke certificate:

certbot revoke --cert-path /etc/letsencrypt/archive/${YOUR_DOMAIN}/cert1.pem

Check certificate's expiry date:

cat /etc/letsencrypt/live/domain.tld/cert.pem | openssl x509 -noout -enddate

GIT

To reset your git repository to given commit id, do:

git reset --hard <commit-id>
git push origin master --force

Ignoring files that are already tracked

git update-index --assume-unchanged <your file here>

Delta Chat Mail sieve

require ["fileinto"];
# rule:[DeltaChat]
if header :contains "Chat-Version" "1.0"
{
	fileinto "DeltaChat";
	stop;
}

Postfix mail queue

To view postfix mail queue in case of problems and remove a particular message from it, do as follows:

# mailq

-Queue ID-  --Size-- ----Arrival Time---- -Sender/Recipient-------
6DCF32201B*    4824 Thu Oct 19 22:54:44  roughnecks@woodpeckersnest.eu
                                         debian@spacenet.it

-- 4 Kbytes in 1 Request.
# postsuper -d 6DCF32201B

postsuper: 6DCF32201B: removed
postsuper: Deleted: 1 message

Check Preferred Outgoing IP (when multiple are set on <interface>)

curl ifconfig.me

The correct command to add a new Linux user (in this case without a shell)

adduser --shell /usr/sbin/nologin <username>

I always forget which command is the complete one, useradd or adduser

Setting up IPv6 in Windows

Written by Simone

Found this post quite useful to set up Hurricane Electric's IPv6 on a Windows machine and serve the connection over LAN (Windows machine must stay on, obviously)

Legend of variables:
$ipv4a   = tunnel server's IPv4 IP
$ipv4b   = user's IPv4 IP
$ipv6a   = tunnel server's side of point-to-point /64 allocation
$ipv6b   = user's side of point-to-point /64 allocation
$ipv6c   = first address of user's routed /64 allocation
$ipv6d   = user's routed /64 allocation
$adapter = name of local area network connection (where your IPv4 address is configured now)

netsh int teredo set state disabled
netsh int ipv6 add v6v4tunnel IP6Tunnel $ipv4b $ipv4a
netsh int ipv6 add address IP6Tunnel $ipv6b

netsh int ipv6 set interface IP6Tunnel forwarding=enabled
netsh int ipv6 set interface "$adapter" forwarding=enabled advertise=enabled
netsh int ipv6 add address "$adapter" $ipv6c
netsh int ipv6 set route $ipv6d "$adapter" publish=yes
netsh int ipv6 add route ::/0 IP6Tunnel $ipv6c publish=yes

Dendrite (Matrix) is live

Written by Simone

Several months ago I was running a "matrix-docker-ansible" playbook on OVH, but limited resources led me to take it down.

More than one time I thought about reliving that server (which I have backups of), but reading on the changes that the playbook got so far and the difficulties in running a Matrix environment without exposing web ports (80 and 443), I thought to better give up.

Yesterday I looked into Dendrite (a Matrix server written in GO), which looks a lot simpler than the whole lot the docker-ansible playbook offers, so I went ahead and installed it the manual way with nginx as reverse proxy. Easy peasy - documentation is quite good.

Today I also installed "Cinny" as a web client and I'm done for the moment... Federation works, registrations are closed (can invite friends though) and apart from RAM peaks and quite long waiting times when joining rooms (up to 2K members), the server is doing quite well: started with 700MB of RAM (at joining time) and went down to less than 300 in normal usage. CPU gets high spikes when sending messages but other than that is just a few points %. Just don't even think to join the official #matrix room or shit will happen 😀

My new Matrix username is: @roughnecks:woodpeckersnest.space, feel free to add me.

Migrate whole system from one host to another

Written by Simone

I recently moved from OVH to Contabo for my VPS: since setting up everything from scratch looked like a hard challenge, I moved everything using rsync.

Firstly, some preparations:

  • installed the same kernel I had on OVH
  • rebooted Contabo
  • installed rsync on Contabo

Then I switched to the OVH shell:

  • created a exclude.txt file with all the directory and file I didn't want to move:
/boot/
/etc/modules
/etc/fstab
/etc/mtab
/etc/netconfig
/etc/networks
/etc/network/
/etc/ssh/
/etc/cloud/
/etc/cloud-release
/etc/grub.d/
/etc/host.conf
/etc/hostname
/etc/hosts
/etc/hosts.allow
/etc/hosts.deny
/etc/init.d/cloud-config
/etc/init.d/cloud-final
/etc/init.d/cloud-init
/etc/init.d/cloud-init-local
/etc/initramfs-tools/
/etc/default/grub
/etc/default/grub.d/
/etc/kernel/
/etc/kernel-img.conf
/lib/modules/
/lost+found/
/sys/
/proc/
/dev/
/tmp/
/var/cache/
/var/log/journal/
/mnt/
/root/.ssh/
  • moved it to /root/exclude.txt
  • stopped all the running services
systemctl stop <service>
  • now we can begin rsync-ing: you have to have access to the root user on destination VPS
rsync -avzP --exclude-from=/root/exclude.txt / root@CONTABO_IP:/
  • after that, I edited /etc/fstab to add support for quota, like I had on OVH
  • then I searched and grepped /etc for my OVH IP address, because I knew it was set somewhere on some config file
grep -r “OVH IP” /etc/*
  • and substituted it with the new Contabo IP, where necessary.

Finally, I could reboot Contabo:

reboot

Once up & running again, I changed all DNS entries from OVH to Contabo IP.. And waited 😀

  • Last but not least, I edited etc/hosts manually to reflect the new Domain Name address and also set up the hostname
hostnamectl set-hostname <new-hostname>

Done!

EDIT! A detail I missed to mention is that your destination host's root password will be changed, after rsync-ing, to the one of the source host!!

conversejs community plugins

Written by Simone

Thanks to Zash and Jcbrand in the "Converse" MUC and a bit of hacking I was able to set up a few community plugins for my conversejs install as a prosody module.

Here's the configuration in /etc/prosody/prosody.cfg.lua:

conversejs_resources = "/usr/local/lib/prosody/modules/mod_conversejs/dist"
conversejs_tags = {
        -- Load favicon
        [[<link rel="shortcut icon" href="https://woodpeckersnest.space/images/converse-js.ico">]];
        -- Load libsignal-protocol.js for OMEMO support (GPLv3; be aware of licence implications)
        [[<script src="https://cdn.conversejs.org/3rdparty/libsignal-protocol.min.js"></script>]];
        -- Load community plugins
        [[<link type="text/css" rel="stylesheet" media="screen" href="conversejs/dist/plugins/search/search.css" />]];
        [[<script src="conversejs/dist/plugins/actions/actions.js"></script>]];
        [[<script src="conversejs/dist/plugins/search/search.js"></script>]];
        [[<script src="conversejs/dist/plugins/search/jspdf.debug.js"></script>]];
        [[<script src="conversejs/dist/plugins/search/jspdf.plugin.autotable.js"></script>]];
        [[<script src="conversejs/dist/plugins/toolbar-utilities/toolbar-utilities.js"></script>]];
        [[<script src="conversejs/dist/plugins/screencast/screencast.js"></script>]];
}

conversejs_options = {
        locked_domain = "woodpeckersnest.space";
        auto_focus = true;
        view_mode = "fullscreen";
        allow_registration = false;
        auto_reconnect = true;
        reuse_scram_keys = true;
        muc_clear_messages_on_leave = true;
        clear_cache_on_logout = false;
        play_sounds = true;
        whitelisted_plugins = {"actions", "search", "toolbar-utilities", "screencast"};
}

You'll have to copy the plugins directories (actions, search etc..) in this path:

/usr/local/lib/prosody/modules/mod_conversejs/dist/plugins/

Then reload configuration and conversejs module or restart prosody.

Already found a bug in "toolbar-utilities" and haven't still had a chance to try the screencast plugin, but they look good for the most part.

Maybe I will add Jitsi Meet or Voice Chat at some point.. Not now though. ¹

EDIT: screencast is working alright, but not in the way you'd expect it. It's not a live streaming, instead it's a recording of your screen which gets uploaded once you stop the cast.. I wouldn't say it's perfect but not even bad.

¹ I've added them 😛

Special characters in DBs are a no-go

Written by Simone

Remember: don't ever use special characters in your DB's users password!!

I have been troubleshooting a migration from mariadbd to pgsql for several hours, until I tried changing my password, which, weirdly enough, was working to connect roundcube but gave errors with pgloader. Once I successfully migrated using the same password with those special characters, roundcube wouldn't connect with pgsql, so I made another fighting round! Finally changed my pgsql password to some letters and numbers only and BAM!, it worked.

Then I stopped mariadbd and disabled the service; now I have more than 200MB RAM freed¹.

¹ See previous post.

mariadbd is a memory hog

Written by Simone

A fat pig eating with its mouth and nose all dirty

I'm still struggling to understand why mariadbd uses so much memory for just 1 DB and 1 application, which in this case is roundcube.

I tried changing this value to no avail.. If anyone wants to share some thoughts I would appreciate:

[mariadb]
innodb_buffer_pool_size=128M

Leave a comment if you know shit. Thanks

EDIT: some details

OVH VPS, 1 core, 2GB, 20 GB SSD. mariadb-server 1:10.11.4-1~deb12u1. roundcube DB size 2.41MB, 17 tables. RAM taken at process restart ~240MB.

Unauthenticated email from [DOMAIN] is not accepted due to 550-5.7.26 domain's DMARC policy.

Written by Simone

I was enabling Exim on my IONOS VPS to deliver email through a smarthost and encountered the error message in this post's title (unable to send to Gmail users.. It's always them!)

After fiddling a while with DMARK and SPF I reconfigured Exim itself to rewrite sender address, so that emails coming from "spacenest.it" (IONOS domain) were sent as coming from "woodpeckersnest.eu", the smarthost and real email server.

Everything is done via this command:

# dpkg-reconfigure exim4-config

And the resulting configuration in /etc/exim4/update-exim4.conf is:

dc_eximconfig_configtype='smarthost'
dc_other_hostnames='cassandra.spacenest.it;spacenest.it'
dc_local_interfaces='127.0.0.1'
dc_readhost='woodpeckersnest.eu'
dc_relay_domains=''
dc_minimaldns='false'
dc_relay_nets=''
dc_smarthost='pandora.woodpeckersnest.space::587'
CFILEMODE='644'
dc_use_split_config='false'
dc_hide_mailname='true'
dc_mailname_in_oh='true'
dc_localdelivery='mail_spool'

On OVH's postfix I just had to modify this line in /etc/postfix/main.cf:

mynetworks = 127.0.0.0/8 [::ffff:127.0.0.0]/104 [::1]/128 94.143.138.27/32

where that last IP is the IP Address of my IONOS server.