How to fix broken monitoring checks using PhantomJS (TypeError) after Ubuntu upgrade

The Ruby version turned out to be not a problem at all. The plugin itself could still be executed. An strace on the plugin execution revealed how phantomjs was executed:

ck@monitoring:~$ strace -f -s 1000 -e trace=execve /usr/lib/nagios/plugins/check_http_load_time.rb -p /usr/local/bin/phantomjs -u https://myapp.example.com/#webview/62
execve("/usr/lib/nagios/plugins/check_http_load_time.rb", ["/usr/lib/nagios/plugins/check_http_load_time.rb", "-p", "/usr/local/bin/phantomjs", "-u", "https://myapp.example.com/#webview/62"], 0x7ffd70e299d8 /* 34 vars */) = 0
execve("/home/admck/bin/ruby", ["ruby", "/usr/lib/nagios/plugins/check_http_load_time.rb", "-p", "/usr/local/bin/phantomjs", "-u", "https://myapp.example.com/#webview/62"], 0x7ffd4f514fb8 /* 34 vars */) = -1 ENOENT (No such file or directory)
execve("/home/admck/.local/bin/ruby", ["ruby", "/usr/lib/nagios/plugins/check_http_load_time.rb", "-p", "/usr/local/bin/phantomjs", "-u", "https://myapp.example.com/#webview/62"], 0x7ffd4f514fb8 /* 34 vars */) = -1 ENOENT (No such file or directory)
execve("/usr/local/sbin/ruby", ["ruby", "/usr/lib/nagios/plugins/check_http_load_time.rb", "-p", "/usr/local/bin/phantomjs", "-u", "https://myapp.example.com/#webview/62"], 0x7ffd4f514fb8 /* 34 vars */) = -1 ENOENT (No such file or directory)
execve("/usr/local/bin/ruby", ["ruby", "/usr/lib/nagios/plugins/check_http_load_time.rb", "-p", "/usr/local/bin/phantomjs", "-u", "https://myapp.example.com/#webview/62"], 0x7ffd4f514fb8 /* 34 vars */) = -1 ENOENT (No such file or directory)
execve("/usr/sbin/ruby", ["ruby", "/usr/lib/nagios/plugins/check_http_load_time.rb", "-p", "/usr/local/bin/phantomjs", "-u", "https://myapp.example.com/#webview/62"], 0x7ffd4f514fb8 /* 34 vars */) = -1 ENOENT (No such file or directory)
execve("/usr/bin/ruby", ["ruby", "/usr/lib/nagios/plugins/check_http_load_time.rb", "-p", "/usr/local/bin/phantomjs", "-u", "https://myapp.example.com/#webview/62"], 0x7ffd4f514fb8 /* 34 vars */) = 0
strace: Process 1371139 attached
strace: Process 1371140 attached
[pid 1371140] execve("/bin/sh", ["sh", "-c", "/usr/local/bin/phantomjs --load-images=yes --local-to-remote-url-access=yes --disk-cache=no  /usr/lib/nagios/plugins/netsniff.js https://myapp.example.com/#webview/62 'false' 2> /dev/null"], 0x56026c5dfba0 /* 34 vars */) = 0
strace: Process 1371141 attached
[pid 1371141] execve("/usr/local/bin/phantomjs", ["/usr/local/bin/phantomjs", "--load-images=yes", "--local-to-remote-url-access=yes", "--disk-cache=no", "/usr/lib/nagios/plugins/netsniff.js", "https://myapp.example.com/#webview/62", "false"], 0x56123391cc78 /* 34 vars */) = 0
[pid 1371141] +++ exited with 1 +++
[pid 1371140] --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=1371141, si_uid=901, si_status=1, si_utime=0, si_stime=1} ---
[pid 1371140] +++ exited with 1 +++
[pid 1371138] --- SIGCHLD {si_signo=SIGCHLD, si_code=CLD_EXITED, si_pid=1371140, si_uid=901, si_status=1, si_utime=0, si_stime=0} ---
UNKNOWN: Could not parse JSON from phantomjs
[pid 1371139] +++ exited with 3 +++
+++ exited with 3 +++

strace also shows the the child process (the phantomjs command) exited with a non-zero exit code. What is happening there?

Let's find out by executing the exact same command:

ck@monitoring:~$ /usr/local/bin/phantomjs --load-images=yes --local-to-remote-url-access=yes --disk-cache=no /usr/lib/nagios/plugins/netsniff.js https://myapp.example.com/#webview/62
Auto configuration failed
140551518091200:error:25066067:DSO support routines:DLFCN_LOAD:could not load the shared library:dso_dlfcn.c:185:filename(libproviders.so): libproviders.so: cannot open shared object file: No such file or directory
140551518091200:error:25070067:DSO support routines:DSO_load:could not load the shared library:dso_lib.c:244:
140551518091200:error:0E07506E:configuration file routines:MODULE_LOAD_DSO:error loading dso:conf_mod.c:285:module=providers, path=providers
140551518091200:error:0E076071:configuration file routines:MODULE_RUN:unknown module name:conf_mod.c:222:module=providers

Yep, that definitely looks like a problem. But does it mean that manually installed phantomjs binary simply won't work anymore on the upgraded Ubuntu 22.04 or later? Or is it something else?

TypeError on packaged phantomjs binary

I tried the Ubuntu packaged phantomjs, located under /usr/bin/phantomjs. This package was still installed, although the package doesn't exist any longer in Ubuntu 22.04.

To make the question mark above my head even bigger, it turns out this binary somewhat worked:

ck@monitoring:~$ /usr/bin/phantomjs --load-images=yes --local-to-remote-url-access=yes --disk-cache=no /usr/lib/nagios/plugins/netsniff.js https://myapp.example.com/#webview/62
{
    "log": {
        "version": "1.2",
        "creator": {
            "name": "PhantomJS",
            "version": "2.1.1"
        },
        "pages": [
            {
                "startedDateTime": "2025-04-15T11:46:40.350Z",
                "endedDateTime": "2025-04-15T11:46:40.546Z",
                "initialResourceLoadTime": 54,
                "id": "https://myapp.example.com/#webview/62",
                "size": 29861,
                "resourcesCount": 4,
                "domElementsCount": 32,
                "title": "Einen Moment bitte, die Ausgabe wird geladen...",
                "jscheckout": null,
                "pageTimings": {}
            }
        ]
    }
}
TypeError: Attempting to change the setter of an unconfigurable property.
TypeError: Attempting to change the setter of an unconfigurable property.

The phantomjs output was clearly able to open the JavaScript elements on the target URL. However the output at the end contains two TypeError lines, making the whole output an invalid JSON.

Now the UNKNOWN error from the monitoring plugin makes sense: Could not parse JSON from phantomjs. Indeed.

Default OPENSSL_CONF causing problems

A good old pre-AI-times Internet research for this PhantomJS TypeError led me to a question on StackOverflow. The same two TypeError lines show up in the OP's question from 2020.

A comment, added 3 years later, mentioned that it was a problem with the SSL configuration:

The way I solved it was: export OPENSSL_CONF=/dev/null

I didn't really think that would solve anything, but at the same time I've also ran into big OpenSSL communication issues after an OS/OpenSSL upgrade in the past.

To turn my question mark above my head even bigger, /usr/bin/phantomjs would now return an even different error.

HOWEVER the manually installed binary at /usr/local/bin/phantomjs would now suddenly work!

ck@monitoring:~$ export OPENSSL_CONF=/dev/null
ck@monitoring:~$ /usr/local/bin/phantomjs --load-images=yes --local-to-remote-url-access=yes --disk-cache=no /usr/lib/nagios/plugins/netsniff.js https://myapp.example.com/#webview/62
{
    "log": {
        "version": "1.2",
        "creator": {
            "name": "PhantomJS",
            "version": "2.1.1"
        },
        "pages": [
            {
                "startedDateTime": "2025-04-15T13:56:09.373Z",
                "endedDateTime": "2025-04-15T13:56:09.609Z",
                "initialResourceLoadTime": 58,
                "id": "https://myapp.example.com/#webview/62",
                "size": 26643,
                "resourcesCount": 4,
                "domElementsCount": 32,
                "title": "Einen Moment bitte, die Ausgabe wird geladen...",
                "jscheckout": null,
                "pageTimings": {}
            }
        ]
    }
}

No error, no TypeError lines at the end; a valid JSON output!

How would this now show when using the check_http_load_time plugin?

ck@monitoring:~$ /usr/lib/nagios/plugins/check_http_load_time.rb -u "https://myapp.example.com/#webview/62" -p /usr/local/bin/phantomjs
OK: https://myapp.example.com/#webview/62 load time: 0.21

Hurray! The monitoring plugin works again with the OPENSSL_CONF environment variable set to something different than the system default (which is /etc/ssl/openssl.cnf, just FYI).

Adjusting Icinga2 CheckCommand to use environment variable

Of course the actual check is executed by Icinga2 and the environment variable is not defined anywhere. It could be set on the "nagios" system user, which is used in our environment to launch checks, however there's a better solution.

The Icinga2 CheckCommand object allows to include environment variables. This is pretty helpful if you need to set environment variables for proxy access or similar in front of a command. In our situation it helps us to set the OPENSSL_CONF environment variable for this specific check.

Here's an excerpt with the relevant part:

root@icingamaster:~# head /etc/icinga2/zones.d/global-templates/commands/check_http_phantom.conf
object CheckCommand "check_http_phantom" {
  import "plugin-check-command"
  command = [ PluginContribDir + "/check_http_load_time.rb" ]
  env.OPENSSL_CONF = "/dev/null"

  arguments = {

After an Icinga2 config reload (systemctl reload icinga2), all the checks using the check_http_load_time plugin turned OK again.

Where to in the future?

With PhantomJS no longer receiving updates or fixes, something else must come in the future, to replace these kind of HTTP checks with JavaScript loaded content. I'm tending towards real E2E testing tools, such as Selenium or Playwright. I've seen a great presentation at the OSMC 2024 about these E2E tools. But right now I have to stick with the PhantomJS based check - due to lack of time :-/.

Anti Spam Question: What's the short name for the United States of America? All capital letters.