Different approach to NVMe subsystem handling #163
Replies: 7 comments 10 replies
-
|
For tracking purposes: #148 Thank you for continuing this dialogue. I will review shortly. |
Beta Was this translation helpful? Give feedback.
-
|
Thinking this through: why couldn't numbers 2 and 5 be together? For example:
|
Beta Was this translation helpful? Give feedback.
-
Because by doing so we would fall back into the old "mixed" approach. Keeping "2" and "5" separate creates a hierarchy between commands: some can only be executed if another has already been executed. As an extreme example, we could combine By keeping all actions separate, we can write smaller functions that perform atomic tasks, and from a user perspective the chain of actions is clear: NQNs are something, hosts are something else, ports are something else, and , once created, these three entities can be related using simple atomic operations. If you recall, during the call we both spent some time staring at the code imagining the various scenarios that could have occurred during the NVMe unexport - and in the end luckily you noticed that yet another check was missing (to avoid deleting a port if used by another NQN). I'd like to divide all the command line activities so as not to have to write convoluted and complex functions forced to consider four/five different scenarios in order not to run into huge errors. Let me know what you think! |
Beta Was this translation helpful? Give feedback.
-
|
@matteotenca OK. You convinced me. Who did you want to do the work? I have no issue doing the implementation but knowing you, it is either already implemented in a branch or part of the way there. :-) |
Beta Was this translation helpful? Give feedback.
-
@pkoutoupis Eheheheheheheheehhe I can do it, if that's okay with you.! This time I did not went straight with the implementation, but I spent some time to realize the autocompletion script for the current shub@ubuserver:~$ source ./rapiddisk_autocomplete.shIf you have time, tell me what you think! :-) |
Beta Was this translation helpful? Give feedback.
-
|
Of course it is okay with me. I just do not want to put any pressure on you. Re: autocompletion tool.... |
Beta Was this translation helpful? Give feedback.
-
#/usr/bin/env bash
# VERSION=0.0.1
# Copyright 2023 Matteo Tenca <[email protected]>
# License: GPL v2
_rapiddisk_completions()
{
local cur prev opts first second third fourth opt1 opt2
new_array=()
# Iterate over the original array
for item in "${COMP_WORDS[@]}"; do
if [[ "$item" != "-j" ]]; then
# Add the item to the new array if it's not "-j"
new_array+=("$item")
fi
done
first="${new_array[1]}"
opt1="${new_array[2]}"
second="${new_array[3]}"
opt2="${new_array[4]}"
third="${new_array[5]}"
opt3="${new_array[6]}"
fourth="${new_array[7]}"
opt4="${new_array[8]}"
# echo -e "\nfirst: '$first'\nsecond: '$second'\nthird: '$third'\nopt1: '$opt1'\nword being completed: '$2'\nprevious: '$prev'\nCOMP_CWORD: '$COMP_CWORD'\nCOMP_WORDS: '${COMP_WORDS[@]}'\ncur: '$cur'\n1: '$1'\n2: '$2'\n3: '$3'"
if [ $COMP_CWORD -eq 1 ] ; then
opts="-j -x -d -h -l -a -r -m -u -f -L -U -i -X -e -R -n -N"
else
case $3 in
-j)
if [ "${COMP_WORDS[1]}" = "-j" ] && [ $COMP_CWORD -eq 2 ] ; then
opts="-x -d -l -a -r -m -u -f -L -U -i -X -e -R -n -N"
fi
;;
-d|-r|-m|-f|-U|-L)
opts=`find /dev -maxdepth 1 -group disk |grep '/dev/rd' | grep -oP 'rd\d+'`
;;
-u)
opts=`ls /dev/mapper/rc-*| sed 's,/dev/mapper/,,g'`
;;
-R)
opts="-b"
;;
-i)
opts=`ip link|grep -oP '^\d+:\s+[^:]+'|grep -Pv '\slo$'|grep -oP '[^\s]+$'`" NULL"
;;
-b)
if [ "$first" = "-e" ] || [ "$first" = "-x" ] || [ "$first" = "-R" ] ; then
opts=`find /dev -maxdepth 1 -group disk |grep '/dev/rd' | grep -oP 'rd\d+'`
else
opts1=`find /dev -maxdepth 1 -group disk`
opts=${opts1}' '`ls /dev/mapper/rc-*`
fi
;;
-p)
opts="wt wa wb"
;;
-t)
if [ "$opt1" != "NULL" ] ; then
opts="tcp rdma"
else
opts="loop"
fi
;;
-X)
opts="-P"
;;
-e|-x)
opts="-b -P -H"
;;
-c)
;;
*)
if [ -n "$first" ] ; then
case $first in
-r)
opts="-c"
;;
-m)
if [ -z "$third" ] || [ "$third" = "-" ] ; then
case $second in
-b)
if [ "$2" = "-b" ] ; then opts="-b" ; else opts="-p" ; fi
;;
-p)
if [ "$2" = "-p" ] ; then opts="-p" ; else opts="-b" ; fi
;;
*)
opts="-b -p"
;;
esac
fi
;;
-i)
if [ -z "$third" ] || [ "$third" = "-" ] ; then
case $second in
-P)
if [ "$2" = "-P" ] ; then opts="-P" ; elif [ -n "$opt2" ] ; then opts="-t" ; fi
;;
-t)
if [ "$2" = "-t" ] ; then opts="-t" ; else opts="-P" ; fi
;;
*)
opts="-P -t"
;;
esac
fi
;;
-e|-x)
if [ -z "$opt2" ] || [ "$opt2" = "-" ] ; then
case $opt1 in
-P)
if [ "$2" = "-P" ] ; then opts="-P" ; elif [ -n "$second" ] ; then opts="-H -b" ; fi
;;
-H)
if [ "$2" = "-H" ] ; then opts="-H" ; elif [ -n "$second" ] ; then opts="-P -b" ; fi
;;
-b)
if [ "$2" = "-b" ] ; then opts="-b" ; elif [ -n "$second" ] ; then opts="-P -H " ; fi
;;
*)
;;
esac
fi
if [ -z "$opt3" ] || [ "$opt3" = "-" ] ; then
case $opt2 in
-P)
if [ "$2" = "-P" ] ; then opts="-P" ; elif [ -n "$third" ] ; then if [ "$opt1" = "-H" ] ; then opts="-b" ; else opts="-H"; fi ; fi
;;
-H)
if [ "$2" = "-H" ] ; then opts="-H" ; elif [ -n "$third" ] ; then if [ "$opt1" = "-P" ] ; then opts="-b" ; else opts="-P"; fi ; fi
;;
-b)
if [ "$2" = "-b" ] ; then opts="-b" ; elif [ -n "$third" ] ; then if [ "$opt1" = "-H" ] ; then opts="-P" ; else opts="-H"; fi ; fi
;;
*)
;;
esac
fi
;;
*)
;;
esac
fi
;;
esac
fi
COMPREPLY=()
cur="${COMP_WORDS[COMP_CWORD]}"
prev="${COMP_WORDS[COMP_CWORD-1]}"
COMPREPLY=( $(compgen -W "$opts" -- ${cur}) )
return 0
}
complete -F _rapiddisk_completions rapiddisk
|
Beta Was this translation helpful? Give feedback.
-
|
@matteotenca I like this. I played with it for a little bit. I look forward to when you open up a PR. Some thoughts on location: And then create a new |
Beta Was this translation helpful? Give feedback.
-
|
I believe it would be fine if the autocompletion file is installed via |
Beta Was this translation helpful? Give feedback.
-
|
I do not mind having it installed by |
Beta Was this translation helpful? Give feedback.
-
|
@pkoutoupis Here is a very first implementation: |
Beta Was this translation helpful? Give feedback.
-
|
I was on a rush, here are some more info:
Let me know what you think! :-) |
Beta Was this translation helpful? Give feedback.
-
|
@matteotenca Thank you for you hard work and persistence in keeping this project both stable and relevant. I will check it out shortly. |
Beta Was this translation helpful? Give feedback.
-
|
@matteotenca Any way you can create that PR? I would like to start commenting. |
Beta Was this translation helpful? Give feedback.
-
@pkoutoupis I was out of town for a month now, I'll be back this weekend and I will manage to get the job done ASAP. Sorry for the long delay! |
Beta Was this translation helpful? Give feedback.
-
|
No worries and no need to be sorry. I hope that you had a nice and relaxing time away. 😄 |
Beta Was this translation helpful? Give feedback.
-
Hi @pkoutoupis !
I was thinking of a new set of switch to handle the creation/deletion of ports, allowed hosts and NQN, and to link them to each other independently.
A switch to create/remove a port, which require to specify the ip OR the network interface AND the port number AND the protocol on creation, just the port on deletion (since a port can be mapped to an IP only, if I am right).
Examples (switch names are casual):
$ rapiddisk -i -P port_number -i interface|ip_address -t tcp|rdma|loop$ rapiddisk -X -P port_numberA switch to add/remove allowed hostname/s globally
$ rapiddisk -y -H hostname[,hostname,..]$ rapiddisk -Y -H hostname[,hostname,..]A switch to add/remove a NQN without linking it to any port/allowed host/s
$ rapiddisk -e -b ramdisk_name$ rapiddisk -x -b ramdisk_nameA switch to link/unlink an existing port to an existing NQN:
$ rapiddisk -z -P port[,port,..] -b ramdisk_name$ rapiddisk -Z -P port[,port,..] -b ramdisk_nameA switch to link/unlink an existing allowed host/s to an existing NQN
$ rapiddisk -w -H hostname[,hostname,..] -b ramdisk_name$ rapiddisk -W -H hostname[,hostname,..] -b ramdisk_nameWe can decide what to do when the user tries to remove an NQN/port/hosts with linked NQN/ports/hosts (for example we can add a force switch which will unlink a NQN from all the used resources and then remove the NQN itself, or just return some information about the current configuration).
Tell me your thoughts
Cheers!!
Teo
Beta Was this translation helpful? Give feedback.
All reactions