Skip to content

Unique bash history on different HPC resources

NCAR's HPC resource share a common home files system, which greatly simplifies working across systems. One slight annoyance I noted with this setup, however, is a common bash history across all machines. While the commands I usually run on Casper and Derecho are very similar, they are not identical. Fortunately, bash allows its history behavior to be customized easily, and with just a little effort we can get HPC resource-specific history implementations.

Customizing your bash history behavior

By default, bash stores its history in the file ~/.bash_history, so shells on different machines will share the same history file. It's really easy to change this behavior with just a little shell customization.

The following snippet is lifted from my ~/.profile, where I do all my shell customization:

Unique HISTFILE per NCAR resource type

~/.profile
 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
# increase the number of lines stored in our history file
# (https://www.gnu.org/software/bash/manual/html_node/Bash-History-Facilities.html)
export HISTSIZE=1024

# customize history behavior
# (https://www.bashsupport.com/bash/variables/bash/histcontrol/)
export HISTCONTROL=ignorespace:erasedups

# Machine-specific history file
case "x${NCAR_HOST}" in
    "x") ;;
    *) export HISTFILE=$HOME/.bash_history.${NCAR_HOST}; ;;
esac

The first two export statements modify the default behavior of the history built-in command, as described in the references.

The case switch statement on lines 10-13 examines the value of the NCAR_HOST environment variable and (if set) uses it to create a unique HISTFILE. NCAR_HOST is set by all our ncarenv modules, so will exist on Derecho and Casper login and compute nodes.

With the configuration above, all my casper and derecho commands stay put and each machine has its own distinct history!

Bonus: search behavior customization

bash used the GNU readline library for interacting with the user, which allows for a lot of behavior customization. One feature I found that I like to use under bash is:

  • history-search-forward ()

    Search forward through the history for the string of characters between the start of the current line and the point. The search string must match at the beginning of a history line.

  • history-search-backward ()

    Search backward through the history for the string of characters between the start of the current line and the point. The search string must match at the beginning of a history line.

which can be enabled via:

~/.profile
# substring search history
bind '"\e[A"':history-search-backward
bind '"\e[B"':history-search-forward

What this means is, say I type $ qs at a terminal prompt and then hit the "up" arrow, my shell will cycle through my history for any commands beginning with qs*, e.g. qsub, qstat.

For other options (and there are many...), see this page.