[Question] About the operation when the pcs command is wrapped in perl etc..

[HideoYamauchi]

Hi All

We have implemented a perl wrapper to make it easier to use pcs on RHEL9.1.

#!/usr/bin/perl
sub exec_command {
    my ($command) = @_;
    my $result;
    my $exit_code;
    $result = `$command`;
    $exit_code = $? >> 8;
    if ($exit_code != 0){
        printf("ERROR\n");
    }
    return $result;
}

my $input = <STDIN>;
chomp $input;
if ($input !~ m/^y$/i) {
    exit(0);
}

close STDIN;

exec_command("/usr/sbin/pcs cluster destroy --all --force");

The pcs command will generate an error if STDIN is closed on the perl side when executing wrap processing with confirmation processing such as the following.

[root@rh91-01 ~]# cat /etc/redhat-release 
Red Hat Enterprise Linux release 9.1 (Plow)
[root@rh91-01 ~]# pcs --version
0.11.3

[root@rh91-01 ~]# ./test.pl
y
Traceback (most recent call last):
  File "/usr/sbin/pcs", line 33, in <module>
    sys.exit(load_entry_point('pcs==0.11.3', 'console_scripts', 'pcs')())
  File "/usr/lib/python3.9/site-packages/pcs/app.py", line 273, in main
    routing.create_router(cmd_map, [])(
  File "/usr/lib/python3.9/site-packages/pcs/cli/common/routing.py", line 33, in _router
    return cmd_map[sub_cmd](lib, argv_next, modifiers)
  File "/usr/lib/python3.9/site-packages/pcs/cli/common/routing.py", line 33, in _router
    return cmd_map[sub_cmd](lib, argv_next, modifiers)
  File "/usr/lib/python3.9/site-packages/pcs/cluster.py", line 1266, in cluster_destroy
    if utils.is_run_interactive():
  File "/usr/lib/python3.9/site-packages/pcs/utils.py", line 2059, in is_run_interactive
    return sys.stdin.isatty() and sys.stdout.isatty()
AttributeError: 'NoneType' object has no attribute 'isatty'
ERROR

Conversely, if STDIN is not closed, the pcs command will not generate an error.

[root@rh91-01 ~]# ./test.pl
y
Warning: Unable to load CIB to get guest and remote nodes from it, those nodes will not be deconfigured.
rh91-02: Stopping Cluster (pacemaker)...
rh91-01: Stopping Cluster (pacemaker)...
rh91-02: Successfully destroyed cluster
rh91-01: Successfully destroyed cluster
[root@rh91-01 ~]# 

I think that the process of closing STDIN used on the perl side is in the normal implementation.

It seems to be a problem of isatty processing on the pcs side, but is it possible to deal with it?

Best Regards,
Hideo Yamauchi.

[tomjelinek]

Hi @HideoYamauchi,

Thank you for reaching out. It seems you found a bug in pcs. Apparently, sys.stdin as well as sys.stdout and sys.stderr can be None. We should update pcs to check if they are not None before we access their isatty method.

[HideoYamauchi]

Hi Tomas,

Thank you for reaching out. It seems you found a bug in pcs. Apparently, sys.stdin as well as sys.stdout and sys.stderr can be None. We should update pcs to check if they are not None before we access their isatty method.

Thanks for your comment.

I understand that there is a problem with pcs.
For the time being, I will consider some kind of correspondence in the wrapping process.

Many thaks,
Hideo Yamauchi.,

[HideoYamauchi]

Hi All,

Thanks for the correction.

Best Regards,
Hideo Yamauchi.