Download presentation
Presentation is loading. Please wait.
Published by尉 惠 Modified over 7 years ago
2
Passwords Passwords are unpleasant Hard to remember Remember a couple
Single points of failure Algorithmic passwords are also flawed Use 'pwgen' to make decent passwords Still hard to remember
3
Storing passwords in a vault
Why not put your passwords in a file Password protect the file Single point of failure has shifted to a private file Still not ideal, but certainly safer than 1 password to each thing Plenty of password managers out there: 1Password, KeePass, LastPass, pass
4
pass password manager Uses GPG keypair for encryption/decryption of passwords Stores each password in a separate file All files kept in a GIT repo Allows push/pulling to/from machines you control Do not put them on GitHub/BitBucket/etc. Lots of platforms supported
5
Making GPG keys – Quick redux
Check LPI-102 Topic 110 gpg --gen-key [Generate a key interactively] gpg --output revoke.asc --gen-revoke mykey [Generate revocation cert. Critical for identity keys!] gpg --export-secret-key -a “MyUser” -o priv.key Export secret key, needed on every device to decrypt passwords gpg --export-public-key -a “MyUser” -o pub.key Export public key, needed on every device to encrypt passwords
6
Setting up pass pass init foo@bar.com Uses GPG key for foo@bar.com
pass git init pass git remote add origin myVPS.com:somefolder pass git push -u all pass generate Logins/Netsoc 32 Makes a 32-char long password in the file Logins/Netsoc pass git push
7
Cool things Each folder (e.g. Logins/Netsoc) can have its own GPG key to en/decrypt $folder/.gpg-id contains the GPG secret key ID The path is, by default, ~/.password-store Use gpg-agent to stop having to enter password repeatedly eval “$(gpg-agent –daemon)” Has clipboard feature; stores password for 45 seconds pass -c Logins/Netsoc
8
Managing lots of machines
Very relevant to the ~~~cloud~~~ Two types of methods: Configuration management Remote shells Lots of different configuration management software: Salt, Ansible, Puppet, Chef, CFEngine, Bcfg2, Spacewalk, etc. Many different approaches too: Enforcing/Passive Agent/Agentless Symmetric/Asymmetric communication layer Config storage: JSON/XML/LDAP/????
9
Remote shells Quick way to execute commands on many servers Fast hack:
tmux with 'setw synchronize-panes' Better way: pdsh fab Also a deployment system
10
Monitoring machines Lots of machines => lots of possible failures
Need monitoring and alerting systems Very detailed monitoring can be done with collectd Can output to syslog, files, CSV, etc. Syslog can be directed to a syslog server (or something like logstash) Round-robin remote monitoring done with cacti/munin Munin is very simple, uses telnet(!) to gather data
12
Monitoring machines A few options Nagios Zabbix
Icinga (holy war with nagios) Nagios uses s/pagers to alert people to problems Problems are when some script nagios fires fails Three states: Normal (or recovered) Warning Critical
14
Nagios Two ways of monitoring:
Get central node to check remote with commands over SSH Generally called “Push monitoring” Ask remote nodes for information using NRPE Generally called “Pull monitoring” NRPE – Nagios Remote Plugin Execution Scripts for testing system exist on client, and are fired when client receives instruction from master Can get mixture of both (if you want to make people sad)
15
Asset management Managing equipment is messy Lots of bureaucracy
Use ralph ( Geared towards physical datacentres Less appropriate for cloud stuff
16
Backups Critical Drives die, and you can be stuck without anything
Lots of software for backups Acronis, Bacula, Amanda, rsync(!) Netsoc uses bacula 4 components: Director: Tells things to backup Storage-daemon: Place where backups are put File-daemon: Places where backups come from Console: Lets you check things
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.