ホーム エクスプローラ ヘルプ
登録 サインイン
jeremy
/
HON95-wiki
同期ミラー https://github.com/HON95/wiki
1
0
フォーク 0
ファイル 課題 0 Wiki
ツリー: 7a87d7f061
ブランチ タグ
HON95-wiki
/
config
/
linux-server
/
applications.md

applications.md 41 KB
履歴 Raw

title: Linux Server Applications breadcrumbs:

  • title: Configuration
  • title: Linux Server --- {% include header.md %}

Using

{:.no_toc}

  • Debian 10 Buster

Apache

Outdated and missing information

Setup

  1. Install: apt install apache2

  2. Update security.conf:

    ServerTokens Prod
ServerSignature Off

Usage

  • Enable/disable stuff: a2<en|dis><conf|mod|site> <...>
  • Test configuration: apache2ctl

Apticron

Sends an emails when APT updates are available.

Setup

  1. Prerequesites:
    • Setup Postfix or similar so the system can actually send mail.
    • Make sure the root email alias is set appropriately.
  2. Install: apt install apticron
  3. Setup the config: /etc/apticron/apticron.conf
    • Create it: cp /usr/lib/apticron/apticron.conf /etc/apticron/apticron.conf
    • The defaults are typically fine.
  4. Modify the check interval in /etc/cron.d/apticron (e.g. 30 23 * * *).
  5. Fix a bug causing it to ignore IPADDRESSNUM and always print all IP adresses:
    1. Open /usr/sbin/apticron.
    2. Find this line: IPADDRESSES=`(echo $( /bin/hostname --all-ip-addresses ) ;
    3. Change it to: IPADDRESSES=`(
  6. Test it: apticron

Avahi Daemon

TODO

Setup

  1. Install: apt install avahi-daemon

AWS CLI

Possibly outdated

Setup

Usage

  • Login: aws configure [--profile <profile>]
    • This will store the credentials for the current Linux user.
    • London region: eu-west-2
    • Output format: json
  • Examples:
    • Upload file: aws s3 cp <local_file> s3://<bucket>/

bitwarden_rs

A free community backend for Bitwarden.

TODO

Ceph

See Storage: Ceph.

Certbot

Setup

  1. Install: apt install certbot
  2. (Optional) Add post-update hook: In /etc/letsencrypt/cli.ini, add renew-hook = systemctl reload nginx or equivalent.

Usage

  • Create using HTTP challenge (auto-renewable): certbot -d <domain> --preferred-challenges=http --webroot --webroot-path=<webroot> certonly
  • Create using DNS channelge (not auto-renewable): certbot -d <domain> --preferred-challenges=dns --manual certonly
  • Dry-run renew: certbot renew --dry-run [--staging]
  • Revoke certificate: certbot revoke --cert-path <cert>

DDNS

Cloudflare

  • Cloudflare does not allow limiting the scope for API keys to specific subdomains, so the key will have access to the whole domain (based on how it's registered).
  • Use e.g. cloudflare-ddns-updater.sh.

Docker

Setup

  1. Install: Docker Documentation: Get Docker Engine - Community for Debian.
  2. (Optional) Setup swap limit:
    • If docker info contains WARNING: No swap limit support, it's not working and should maybe be fixed.
    • It incurs a small performance degredation and is optional but recommended.
    • In /etc/default/grub, add cgroup_enable=memory swapaccount=1 to GRUB_CMDLINE_LINUX.
    • Run update-grub and reboot.
  3. Configure /etc/docker/daemon.json:
    • Enable IPv6: "ipv6": true and "fixed-cidr-v6": "<ipv6-subnet>/64"
    • Set DNS servers: "dns": ["1.1.1.1", "2606:4700:4700::1111"]
      • If not set, containers will use 8.8.8.8 and 8.8.4.4 by default.
      • /etc/resolv.conf is limited to only three name servers, so don't provide too many. One may be set by the container itself.
    • (Not recommended) Disable automatic IPTables rules: "iptables": false
  4. (Not recommended on servers) Allow certain users to use Docker: Add them to the docker group.

Usage

  • Miscellanea:
    • Show disk usage: docker system df -v
  • Cleanup:
    • Prune unused images: docker image prune -a
    • Prune unused volumes: docker volume prune
  • Docker run options:
    • Set name: --name=<name>
    • Run in detatched mode: -d
    • Run using interactive terminal: -it
    • Automatically remove when stopped: --rm
    • Automatically restart: --restart=unless-stopped
    • Use "tini" as entrypoint and use PID 1: --init
    • Set env var: -e <var>=<val>
    • Publish network port: -p <host-port>:<cont-port>[/udp]
    • Mount volume: -v <vol>:<cont-path> (<vol> must have a path prefix like ./ or / if it is a directory and not a named volume)

Miscellanea

  • For automatically updating containers, use e.g. watchtower.
  • For managing containers in a pretty web UI, use e.g. Portainer.

Networking

  • Containers in production should not use the default Docker networks.
  • Try to isolate container communication into as small networks as possible (e.g. one network per group of containers for an application).
  • Docker doesn't integrate with ip6tables at all, meaning certain IPv6 features are lacking. For instance, IPv6 is not NATed like IPv4 and ICC can't be disabled. NAT66 shouldn't generally be used in the first place, but the lack of it means IPv6 requires a bit of extra configuration to get it working with containers. IPv6 routing and port publishing work as they should, though, as they don't use ip6tables.
  • Network types:
    • Bridge: A plain bridge where all containers and the host can communicate. Can optionally be directly connected to a host bridge, but that doesn't always work as expected. Vulnerable to ARP/NDP spoofing.
    • Overlay: Overlay network for swarm stuff.
    • Host: The container use the network stack of the host. Ports are published directly to the host.
    • MACVLAN: Bridges connected to a host (parent) interface, allowing containers to be connected to a network the host is part of. Can optionally use trunking on the host interface. All communication between containers and the host is dropped (consider using a host-connected bridge if you need this).
    • L2 IPVLAN: Similar to MACVLAN, but all containers use the host's MAC address. Containers can communicate, but the host can't communicate with any containers.
    • L3 IPVLAN: Every VM uses a separate subnet and all communication, internally and externally, is routed. Should avoid ARP/NDP spoofing. (TODO: Containers and the host can communicate?)
  • Create:
    • Create bridged network: docker network create --driver=bridge --subnet=<ipv4-net> --ipv6 --subnet=<ipv6-net> <name>
    • Create external bridged network (experimental, doesn't work as intented in some scenarios): docker network create --driver=bridge --subnet=<ipv4-net> --gateway=<ipv4-gateway> --ipv6 --subnet=<ipv6-net> --gateway=<ipv6-gateway> -o "com.docker.network.bridge.name=<host-if> <name>
    • Create MACVLAN: docker network create --driver=macvlan --subnet=<ipv4-net> --gateway=<ipv4-gateway> --ipv6 --subnet=<ipv6-net> --gateway=<ipv6-gateway> -o parent=<netif>[.<vid>] <name>
    • Create L2 IPVLAN with parent interface: docker network create --driver=ipvlan --subnet=<ipv4-net> --gateway=<ipv4-gateway> --ipv6 --subnet=<ipv6-net> --gateway=<ipv6-gateway> -o parent=<netif> <name>
  • Use:
    • Run container with network: docker run --network=<net-name> --ip=<ipv4-addr> --ip6=<ipv6-addr> --dns=<dns-server> [...] <image>

Docker Compose

Setup

  1. Install Docker: See above.
  2. Install: Docker Documentation: Install Docker Compose.
  3. Install command completion: Docker Documentation: Command-line completion.

Troubleshooting

Fix Docker Compose No-Exec Tmp-Dir

Docker Compose will fail to work if /tmp has noexec.

  1. Move /usr/local/bin/docker-compose to /usr/local/bin/docker-compose-normal.
  2. Create /usr/local/bin/docker-compose with the contents below and make it executable.
  3. Create the new TMPDIR dir.

New docker-compose:

#!/bin/bash
# Some dir without noexec
export TMPDIR=/var/lib/docker-compose-tmp
/usr/local/bin/docker-compose-normal "$@"

Fail2ban

Setup

  1. Install fail2ban.
  2. Fix the firewall first so it configures itself correctly wrt. firewall blocking.
  3. Check the status with fail2ban-client status [sshd].

Google Authenticator

Possibly outdated

This setup requires pubkey plus MFA (if configured) plus password.

Setup

  • Warning: Keep a shell open and test with a new shell during the process to make sure you don’t lock yourself out.
  • Install: apt install libpam-google-authenticator
  • In /etc/pam.d/sshd, add auth required pam_google_authenticator.so nullok after @include common-auth.

  • In /etc/ssh/sshd_config, set:

    ChallengeResponseAuthentication yes
UsePAM yes
AuthenticationMethods publickey,keyboard-interactive

  • Restart sshd and check that you can login with pubkey and MFA now.

  • (Optional) Add my google-auth-config-prompter.sh profile script to /etc/profile.d/ to ask user to configure Google Auth on login.

  • To allow a group to use only pubkey (no password or OTP):

    • In /etc/ssh/sshd_config, add Match Group no-mfa containing AuthenticationMethods publickey (indented) at the bottom.
    • Add the system group no-mfa and add special users to it.

  • To manually configure MFA for a user:

    • Example: google-authenticator -tduW
    • Use time-based tokens.
    • Restrict usage of the same token multiple times.
    • Don’t rate limit.
    • Allow 3 concurrent codes (1 before, 1 after).

Intel SSD Data Center Tool (isdct)

See Storage: isdct.

Grafana

Typically used with a data source like Prometheus.

Setup (Docker)

  1. See (Grafana) Run Grafana Docker image.
  2. Mount:
    • Config: ./grafana.ini:/etc/grafana/grafana.ini:ro
    • Data: ./data:/var/lib/grafana/:rw (requires UID 472)
    • Logs: ./logs:/var/log/grafana/:rw (requires UID 472)
  3. Configure grafana.ini.
  4. Open the webpage to configure it.

Notes

  • Be careful with public dashboards. "Viewers" can modify any query and thus query the entire data source for the dashboard, unless you have configured some type of access control for the data source (which you probably haven't).

Home Assistant

See Home Assistant.

ISC DHCP Server

Notes

  • DHCPv4 uses raw sockets, so it bypasses the firewall (i.e. no firewall rules are needed). DHCPv6, however, does not. This includes the respective clients as well.
  • The DHCPv6 server is typically used with radvd for router advertisements.

Setup

  1. Install and enable isc-dhcp-server.
  2. Setup config files:
    • DHCPv4: /etc/dhcp/dhcpd.conf
    • DHCPv6 (optional): /etc/dhcp/dhcpd6.conf
  3. If using systemd-networkd, fix wrong startup order:
    • TODO

Configuration

  • Always specify the authorative statement in subnet declarations so that the server will reply with DHCPNAK for misconfigured clients. This may significantly reduce reconfiguration delay when a client moves between subnets.
  • For range6, prefer using CIDR notation. If using range notation, try to align the start and end on a CIDR block to avoid excessive memory usage.
  • DHCPv6 uses lease pools of 9973 entries, so using range sizes below this number may be preferable as a very general reference. /116 gives 8191 addresses.

lm_sensors

Get sensor values like temperature, voltage, fan speeds, etc.

Setup

  1. Install: apt install lm-sensors
  2. Test run it: sensors
  3. Run sensors-detect. When it asks, add the modules to /etc/modules.
  4. Load new modules: systemctl restart kmod
  5. Test run it: sensors

Troubleshooting

  • There's module/chip/sensor errors in the output or journalctl:
    • If you know which chip and sensor (e.g. if it shows it during output), try adding chip "<chip>"\n ignore <sensor> in /etc/sensors3.conf. Re-run sensors. (See Kernel ACPI Error SMBus/IPMI/GenericSerialBus (ServerAdminBlog) for an example on certain HP servers.)
    • If you know which module it is, try to unload it. Re-run sensors. If it worked, then remove it from /etc/modules to make it permanent.

MariaDB

A MySQL fork that is generally MySQL compatible.

Setup

  1. Install: apt install mariadb-server
  2. Run the initial configuration: mysql_secure_installation
    • Set a new MyriaDB root password.
    • Remove all anmonymous/test stuff.
    • Disallow remote root login.

Usage

  • Open prompt: mariadb [-u <user> [-p]]
    • The default user is root.
    • The password can be entered interactively by specifying -p.
    • A password is typically not needed.
  • Add new admin user: GRANT ALL ON *.* TO '<user>'@'127.0.0.1' IDENTIFIED BY '<password>' WITH GRANT OPTION;

NFS

The instructions below use NFSv4 without Kerberos. This should only be used on trusted networks and requires manual user and group ID management.

Server (without Kerberos)

Setup

  1. (Recommended) Use NTP on both server and clients to make sure the clocks are synchronized.
  2. Install: apt install nfs-kernel-server
    • Install portmap if you need support for NFSv2 and v3 (not NFSv4).

  3. (Recommended) Enable only v4:

    1. In /etc/default/nfs-common, set:

      NEED_STATD="no"
NEED_IDMAPD="yes"

    2. In /etc/default/nfs-kernel-server, set:

      RPCNFSDOPTS="-N 2 -N 3"
RPCMOUNTDOPTS="--manage-gids -N 2 -N 3"

    3. Mask "rpcbind":

      systemctl disable --now rpcbind.service
systemctl mask rpcbind.service
systemctl mask rpcbind.socket

    4. Restart it: systemctl restart nfs-server.service

    5. See which versions are running: cat /proc/fs/nfsd/versions (- means disabled)

Usage

  1. Setup a new directory contain all exports in:
    1. Create the root export containing other export dirs: mkdir /export
    2. Create the export mount dirs within the container.
    3. Mount the exports in the container using bind mounts.
      • Example fstab entry using ZFS: /zfspool/alpha /export/alpha none bind,defaults,nofail,x-systemd.requires=zfs-mount.service 0 0
    4. Remember to set appropriate permissions.
  2. Add filesystems to export in /etc/exports.
    • See the example config below.
    • For a list of options, see exports(5).
  3. Update the NFS table: exportfs -ra
    • Or, restart the service: systemctl restart nfs-server.service
  4. (Optional) Show exports: exportfs -v
  5. (Optional) Update the firewall:
    • NFSv4 uses only TCP port 2049.

Example /etc/exports:

# "fsid=root" is a special root export in NFSv4 where other exports are accessible relative to it.
# "sync" should generally always be used. While "async" gives better performance, it violates the spec and may cause data loss in case of power loss.
# "root_squash" maps client root users to an anon user to prevent remote root access. If that's desired, set "no_root_squash" instead.
# "no_subtree_check" disables subtree checking. Subtree checking may be appropriate for certain file systems, but in general it may cause more problems than it solves.
# "insecure" allows clients connecting from non-well-known ports.
/export/ *(fsid=root,ro,sync,root_squash,no_subtree_check,insecure)
/export/projects/ *(rw,sync,root_squash,no_subtree_check,insecure)

Client (without Kerberos)

Setup

  1. Install: apt install nfs-common

Usage

  1. Create a dir to mount the export to.
  2. (Optional) Try to mount it:
    • Command: mount -t nfs4 <server-hostname>:<export> <mountpoint>
    • Note that for NFSv4 with a root export, the export path is relative to the root export.
  3. (Optional) Make it permanent by adding it to fstab.
    • /etc/fstab entry: <nfs-server>:<export> <local-dir> nfs4 defaults 0 0

ntopng

Setup

  1. Install ntopng.
  2. Make sure service ntopng is enabled and running.
  3. Fix log dir owner: chown nobody:nogroup /var/log/ntopng
  4. Configure:
    1. Open /etc/ntopng.conf.
    2. Add -W=<new_port> to enable HTTPS.
    3. (Optional) Set -w=0 to disable HTTP.
  5. Restart it (takes a while).

ntpd

Setup

  1. Disable systemd-timesyncd NTP client by disabling and stopping systemd-timesyncd.
  2. Install ntp.
  3. In /etc/ntp.conf, replace existing servers/pools with ntp.justervesenet.no with the iburst option.
  4. Test with ntpq -pn (it may take a minute to synchronize).

NUT

Setup

Instructions for both servers and clients. Exclusive steps are marked "(Server)" or "(Client)".

Since SSL/TLS is not enabled by default for client-server communication, use only trusted networks for this communication.

  1. Install: apt install nut
    • The service will fail to start since NUT is not configured yet.
  2. Set the mode: Open /etc/nut/nut.conf and set MODE=netserver for server or MODE=netclient for client.
  3. (Server) Add the UPS(s): Open /etc/nut/ups.conf and add a declaration for all UPSes (see example below).
    • Try using the usbhid-ups driver if using USB. Otherwise, check the hardware compatibility list to find the correct driver. If the exact model isn't there, try a similar one.
    • For usbhid-ups, see the example below and usbhid-ups(8).
    • You may need to modify some udev rules, but probably not.
  4. (Server) Restart driver service: systemctl restart nut-driver.service
  5. (Server) Set up local and remote access: Open /etc/nut/upsd.conf and set LISTEN ::.
    • Alternatively add one or multiple LISTEN directives for only the endpoints you wish to listen on.
  6. (Server) Set up users: Open /etc/nut/upsd.users and add users (see example below).
    • Each client should have a separate user.
  7. (Server) Restart the server service: systemctl restart nut-server.service
  8. (Client) TODO: Something about nut-client.service.
  9. Monitor the UPS: Open /etc/nut/upsmon.conf and add MONITOR <ups>@<host>[:<port>] <ups-count> <user> <password> <master|slave>.
    • ups-count is typically 1. If this system is not powered by the UPS but you want to monitor it without shutting down, set it to 0.
  10. (Optional) Tweak upsmon:
    • Set RBWARNTIME (how often upsmon should complain about batteries needing replacement) to an appropriate value, e.g. 604800 (1 week).
  11. (Optional) Add a notify script to run for certain events:
    • In /etc/nut/upsmon.conf, add EXEC to all NOTIFYFLAG entries you want to run the script for (typically all except LOWBATT).
    • In /etc/nut/upsmon.conf, set the script to run using format NOTIFYCMD /opt/scripts/nut-notify.sh.
    • Create the executable script. See an example below for sending email (if Postfix is set up).
  12. Restart monitoring service: systemctl restart nut-monitor.service
  13. Check the log to make sure nut-monitor successfully connected to the server.
    • Note that upsc does not use a server user or the monitoring service, so it's not very useful for debugging that.
  14. Configure delays:
    1. Figure out how much time is needed to shut down the master and all slaves, with some buffer time.
    2. Set the remaining runtime and remaining battery charge for when the UPS should send the "battery low" event (requires admin login): upsrw -s battery.runtime.low=<seconds> <ups> and upsrw -s battery.charge.low=<percent> <ups>
      • This may not work on all UPSes, even if the values appear to be modifiable. This means you're probably stuck with the defaults.
    3. Set the delay from when the master issues the shutdown command to the UPS, to when the UPS powers off; and the delay from when the UPS receives power again to when it should turn on power: For usbhid-ups, this is set using offdelay and ondelay. Otherwise, it's set using ups.delay.shutdown and ups.delay.start. The start delay must be greater than the stop delay.
      • The shutdown command is issued from the master after it's done waiting for itself and slaves and is shutting itself down. The shutdown delay may be useful to increase if there are slaves that take much longer than the master to shut down.
    4. Restart the affected NUT services.
  15. Simulate a power loss, which should power off all monitoring clients and then the UPS: upsmon -c fsd
    • If the client machines are not given enough time to power off before the UPS powers off, you need to modify the shutdown delay settings in the UPS.

Example USB UPS declaration for usbhid-ups (/etc/nut/ups.conf):

[alpha]
    desc = "PowerWalker VI 3000 RLE"
    # usbhid-ups should work for most UPSes with
    driver = usbhid-ups
    # If you have multiple UPSes connected, see usbhid-ups(8) for more specifying which USB device it should use
    port = auto
    # Sets "ups.delay.shutdown", the delay between the shutdown command and when the UPS powers off (default 20s)
    offdelay = 60
    # Sets "ups.delay.start", which has something to do with letting the UPS charge enough to make sure devices may fully boot (default 30s, must be greater than offdelay)
    ondelay = 120

Example server users (/etc/nut/upsd.users):

[admin]
    password = <password>
    actions = SET
    instcmds = ALL

[local]
    password = <password>
    upsmon master

Example notify script:

#!/bin/bash
echo -e "Time: $(date)\nMessage: $@" | mail -s "NUT: $@" root

OpenSSL

Usage

  • Many OpenSSL default options are insecure and must be specified.
  • Specifying -noout -text prints the data as formatted text instead of raw Base64.
  • Create self-signed cert: openssl req -new -x509 -sha256 -newkey rsa:4096 -nodes -keyout key.pem -out cert.pem -days 3650 -subj "/C=ZZ/ST=Local/L=Local/O=Local/OU=Local/CN=localhost"

Pi-hole (Docker)

  • (Optional) Set up an upstream DNS server.
  • Image: pihole/pihole
  • Run on LAN-accessible bridge.
  • Don’t give capability NET_ADMIN.
  • Add a reject rule in the firewall to more easily block HTTPS ads.
  • Find the admin password with docker logs pihole 2>&1 | grep "random password"
  • Specify the upstream DNS server twice so that it doesn’t choose the second itself.
  • Whitelists and blacklists:

Portainer

Standalone Server Setup

Is typically run on a Docker host. Includes the agent.

  1. docker run -d -p 8000:8000 -p 9000:9000 --name=portainer --restart=always -v /var/run/docker.sock:/var/run/docker.sock -v ./data:/data portainer/portainer:<version>
    • Port 9000 is the web UI.
    • Port 8000 is an SSH tunnel server for communicating with agents.
  2. Open the web UI through port 9000 (by default) or a reverse proxy to configure it.
    • If /var/run/docker.sock was mounted, use "local".

Standalone Agent Setup

Must be run on a Docker host. For extra Docker hosts you want to control with another Portainer server.

  1. docker run -d -p 9001:9001 --name portainer_agent --restart=always -v /var/run/docker.sock:/var/run/docker.sock -v /var/lib/docker/volumes:/var/lib/docker/volumes portainer/agent:<version>
  2. TODO

Postfix

Satellite system

Notes

  • When using an SMTP relay, the original IP address will likely be found in the mail headers.
  • Make sure DNS is configured correctly (SPF, DKIM, DMARC).

Setup

  1. Install: postfix libsasl2-modules mailutils
    • If asked, choose to configure Postfix as a satellite system.
  2. Set the FQDN in /etc/postfix/main.cf.
  3. Update the root alias to point your real email address in /etc/aliases, then run newaliases.
  4. Update the main.cf config (example not provided here).
    1. Only listen to localhost: Set inet_interfaces = loopback-only
    2. Disable relaying: Set mynetworks = 127.0.0.0/8 [::ffff:127.0.0.0]/104 [::1]/128
    3. Anonymize banner: smtpd_banner = $myhostname ESMTP
  5. See the specific relay guides:
  6. Setup address rewrite rules:
    • For fixing the To and From fields, which is typically from root to root.
    • Add the rewrite config (see example below).
    • Reference the config using smtp_header_checks in the main config.
    • Test: postmap -fq "From: root@$(hostname --fqdn)" regexp:smtp_header_checks
  7. Setup relay credentials (SASL):
    1. Credentials file: /etc/postfix/sasl_passwd
    2. Add your credentials using format: [relay_domain]:port user@domain:password
    3. Run: postmap sasl_passwd
    4. Fix permissions: chmod 600 sasl_passwd*
  8. Restart postfix.
  9. Try sending an email: echo "Test from $(hostname) at time $(date)." | mail -s "Test" root

File smtp_header_checks:

/^From:\s*.*\S+@node\.example\.bet.*.*$/ REPLACE From: "Node" <node@example.net>
/^To:\s*.*\S+@node\.example\.net.*$/ REPLACE To: "Someone" <someone@example.net>

Usage

  • Send a test mail: echo "Test from $HOSTNAME at time $(date)." | mail -s "Test" root
  • Test the config: postconf > /dev/null
  • Print the config: postconf -n
  • If mailq tells you mails are stuck in the mail queue because of previous errors, run postqueue -f to flush them.

Prometheus

Typically used with Grafana and sometimes with Cortex/Thanos in-between.

Setup (Docker)

  1. See (Prometheus) Installation.
  2. Set the retention period and size:
    • (Docker) Find and re-specify all default arguments. Check with docker inspect or the source code.
    • Add the command-line argument --storage.tsdb.retention.time=15d and/or --storage.tsdb.retention.size=100GB (with example values).
    • Note that the old storage.local.* and storage.remote.* flags no longer work.
  3. Mount:
    • Config: ./prometheus.yml:/etc/prometheus/prometheus.yml:ro
    • Data: ./data/:/prometheus/:rw
  4. Configure prometheus.yml.
    • I.e. set global variables (like scrape_interval, scrape_timeout and evaluation_interval) and scrape configs.
  5. (Optional) Setup remote storage to replicate all scraped data to a remote backend.
  6. (Optional) Setup Cortex or Thanos for global view, HA and/or long-term storage.

Notes

  • The open port (9090 by default) contains both the dashboard and the query API.
  • You can check the status of scrape jobs in the dashboard.
  • Prometheus does not store data forever, it's meant for short- to mid-term storage.
  • Prometheus should be "physically" close to the apps it's monitoring. For large infrastructures, you should use multiple instances, not one huge global instance.
  • If you need a "global view" (when using multiple instances), long-term storage and (in some way) HA, consider using Cortex or Thanos.
  • Since Prometheus receives an almost continuous stream of telemetry, any restart or crash will cause a gap in the stored data. Therefore you should generally always use some type of HA in production setups.
  • Cardinality is the number of time series. Each unique combination of metrics and key-value label pairs (yes, including the label value) amounts to a new time series. Very high cardinality (i.e. over 100 000 series, number taken from a Splunk presentation from 2019) amounts to significantly reduced performance and increased memory and resource usage, which is also shared by HA peers (fate sharing). Therefore, avoid using valueless labels, add labels only to metrics they belong with, try to limit the numer of unique values of a label and consider splitting metrics to use less labels. Some useful queries to monitor cardinality: sum(scrape_series_added) by (job), sum(scrape_samples_scraped) by (job), prometheus_tsdb_symbol_table_size_bytes, rate(prometheus_tsdb_head_series_created_total[5m]), sum(sum_over_time(scrape_series_added[5m])) by (job). You can also find some useful stats in the dashboard.

About Cortex and Thanos

  • Two similar projects, which both provide global view, HA and long-term storage.
  • Cortex is push-based using Prometheus remote writing, while Thanos is pull-based using Thanos sidecars for all Prometheus instances.
  • Global view: Cortex stores all data internally, while Thanos queries the Prometheus instances.
  • Prometheus HA: Cortex stores one instance of the received data (at write time), while Thanos queries Prometheus instances which have data (at query time). Both approaches removes gaps in the data.
  • Long-term storage: Cortex periodically flushes the NoSQL index and chunks to an external object store, while Thanos uploads TSDB blocks to an object store.

Prometheus Exporters

List of Exporters and Software

This list contains exporters and software with built-in exposed metrics I typically use. Some are described in more detail in separate subsections.

Software with exposed metrics

Exporters

Special

Prometheus Node Exporter

Can be set up either using Docker (prom/node-exporter), using the package manager (prometheus-node-exporter on Debian), or by building it from source. The Docker method provides a small level of protection as it's given only read-only system access. The package version is almost always out of date and is typically not optimal to use. If Docker isn't available and you want the latest version, build it from source.

Setup (Downloaded Binary)

See Building and running.

Details:

  • User: prometheus
  • Binary file: /usr/bin/prometheus-node-exporter
  • Service file: /etc/systemd/system/prometheus-node-exporter.service
  • Configuration file: /etc/default/prometheus-node-exporter
  • Textfile directory: /var/lib/prometheus/node-exporter/

Instructions:

  1. Install requirements: apt install moreutils
  2. Find the link to the latest tarball from the download page.
  3. Download and unzip it: wget <url> and tar xvf <file>
  4. Move the binary to the system: cp node_exporter*/node_exporter /usr/bin/prometheus-node-exporter
  5. Make sure it's runnable: node_exporter -h
  6. Add the user: useradd -r prometheus
    • If you have hidepid setup to hide system process details from normal users, remember to add the user to a group with access to that information. This is only required for some metrics, most of them work fine without this extra access.
  7. Create the required files and directories:
    • touch /etc/default/prometheus-node-exporter
    • mkdir -p /var/lib/prometheus/node-exporter/
  8. Create the systemd service /etc/systemd/system/prometheus-node-exporter.service, see prometheus-node-exporter.service.
  9. (Optional) Configure it:
    • The defaults work fine.
    • File: /etc/default/prometheus-node-exporter
    • Example: ARGS="--collector.processes --collector.interrupts --collector.systemd" (enables more detailed process and interrupt collectors)
  10. Enable and start the service: systemctl enable --now prometheus-node-exporter
  11. (Optional) Setup textfile exporters.

Textfile Collector

Collector Scripts

Some I typically use.

Setup and Usage

  1. Set the collector script output directory using the CLI argument --collector.textfile.directory=<dir>.
    • Example dir: /var/lib/prometheus/node-exporter/
    • If the node exporter was installed as a package, it can be set in the ARGS variable in /etc/default/prometheus-node-exporter.
    • If using Docker, the CLI argument specified as part of the command.
  2. Download the collector scripts and make them executable.
    • Example dir: /opt/prometheus/node-exporter/textfile-collectors/
  3. Add cron jobs for the scripts using sponge to wrote to the output dir.
    • Make sure sponge is installed. For Debian, it's found in the moreutils package.
    • Example cron file: /etc/cron.d/prometheus-node-exporter-textfile-collectors
    • Example cron entry: 0 * * * * root /opt/prometheus/node-exporter/textfile-collectors/apt.sh | sponge /var/lib/prometheus/node-exporter/apt.prom

Prometheus Blackbox Exporter

Monitor Service Availability

Add a HTTP probe job for the services and query for probe success over time.

Example query: avg_over_time(probe_success{job="node"}[1d]) * 100

Monitor for Expiring Certificates

Add a HTTP probe job for the services and query for probe_ssl_earliest_cert_expiry - time().

Example alert rule: probe_ssl_earliest_cert_expiry{job="blackbox"} - time() < 86400 * 30 (30 days)

Pterodactyl

General

  • The panel must be able to communicate with all daemons and all vice versa. The user must be able to communicate with both the panel and daemons.
  • Both the panel and daemons need valid TLS certificates.

Panel (Docker)

Setup

TODO

Logs are located in /app/storage/logs/laravel/ inside the container.

Daemon

  1. Follow the official guide.
  2. Install unzip.
  3. Setup a valid TLS certificate.
  4. Setup Docker DNS servers: Add { "dns": ["1.1.1.1", "1.0.0.1", "2606:4700:4700::1111", "2606:4700:4700::1001"] } to /etc/docker/daemon.json.

Game Servers

General

  • You can typically watch the installation progress by watching the docker logs.

Counter-Strike: Global Offensive

See Counter-Strike: Global Offensive (CS:GO).

Team Fortress 2

See Team Fortress 2 (TF2).

Router Advertisement Daemon (radvd)

Setup

  1. Install and enable radvd.
  2. Setup config file: /etc/radvd.conf

Samba

Server

Setup

  1. Install: apt install samba
  2. Open TCP port 445 (and 139 if using NetBIOS).
  3. (Optional) Disable NetBIOS: systemctl disable --now nmbd and systemctl mask nmbd
  4. Configure it (see usage).

Usage

  • Enforce encryption and signing (server signing and smb encrypt) on important volumes.
  • Performance tuning:
    • Socket options: socket options = TCP_NODELAY SO_KEEPALIVE IPTOS_LOWDELAY
    • If the stuff is not important and the network is secure and high throughput is desired: smb encrypt = disabled
    • Raw IO: read raw = yes and read raw = yes
    • Sendfile: use sendfile = yes
    • Zero-copy from net to FS (doesn't work for signed connections): min receivefile size = 16384
    • Async RW for large files: aio read size = 16384 and aio write size = 16384
  • Making changes:
    • Change the configuration file: /etc/samba/smb.conf
    • Test the configuration: testparm -t
    • Restart the service: systemctl restart smbd
  • Manage access to a share:
    • Add a Linux group for the share, like "smb-media", to restrict user access.
    • Fix permissions for only that group on the system.
    • Configure the share to only allow that group.
    • Add Linux users to the group.
  • Manage users:
    • Samba users are somewhat using Linux users but with a different password.
    • To separate pure Samba users from real users, you can add a "smb-" prefix to its username and make it a system user.
    • Create a new Linux (system) user without shell login: useradd -r <name>
      • Or: useradd
    • Add a user and set its password: smbpasswd -a <user>
    • Show users: sudo pdbedit -L -v

Client

Setup

  1. Install: apt install cifs-utils
  2. Add permanent shares (see usage).

Usage

  • Add permanent share:

    1. Create the mountpoint.

    2. Create a credentials file (/root/.credentials/smb/<whatever>):

      user=<user>
password=<password>

    3. In /etc/fstab, add: //<share> <mountpoint> cifs vers=3.1.1,uid=<uid>,gid=<gid>,credentials=<file>,iocharset=utf8 0 0

    4. Test it: mount -a

smartmontools

  • For monitoring disk health.
  • Install: apt install smartmontools
  • Show all info: smartctl -a <dev>
  • Tests are available in foreground and background mode, where foreground mode is given higher priority.
  • Tests:
    • Short test: Can be useful to quickly identify a faulty drive.
    • Long test: May be used to validate the results found in the short test.
    • Convoyance test: Intended to quickly discover damage incurred during transportation/shipping.
    • Select test: Test only the specified LBAs.
  • Run test: smartctl -t <short|long|conveyance|select> [-C] <dev>
    • -C: Foreground mode.

TFTP-HPA

Setup

  1. Install tftpd-hpa.
  2. Update /etc/default/tftpd-hpa based on the config below.
  3. Create the folder /var/tftp with permissions 777 and user-group pair tftp:tftp.
  4. Restart it.
  5. If it shouldn't automatically start, disble it.

File /etc/default/tftpd-hpa:

TFTP_USERNAME="tftp"
TFTP_DIRECTORY="/var/tftp"
TFTP_ADDRESS=":69"
TFTP_OPTIONS="--create --secure"

Unbound

Setup

  1. Install: apt install unbound dns-root-data
    • It may fail to start due to systemd-resolved listening to the DNS UDP port.
  2. Setup the config: /etc/unbound/unbound.conf
  3. Make sure /etc/hosts contains the short and FQDN hostnames.
  4. Setup systemd-resolved:
    1. Open /etc/resolv.conf.
    2. Set DNSStubListener=no.
    3. Set DNS=::1.
    4. Restart systemd-resolved.
  5. Setup resolv.conf:
    1. Open /etc/resolv.conf.

    2. Set:

      nameserver 127.0.0.1
nameserver ::1
domain <domain>
search <domain-list>
  6. Restart unbound: systemctl restart unbound
  7. Test DNSSEC:
    • drill sigfail.verteiltesysteme.net should give an rcode of SERVFAIL.
    • drill sigok.verteiltesysteme.net should give an rcode of NOERROR.
  8. Make sure dns-root-data is updating root hints in file /usr/share/dns/root.hints.

Notes

  • Use DNS over HTTPS/TLS.
  • Setup private addresses for DNS rebinding prevention.

UniFi

See Ubiquiti UniFi Controller (Debian).

ZFS

See Storage: ZFS.

{% include footer.md %}