Make current version test more robust

[m4z]

As mentioned in the margin in #33, we often see the "current_rocketchat_version" check fail (on Debian 9.13) when running "rocketchatctl check-updates" and "rocketchatctl update", and that seems to cause additional problems, like in this case (first without debug output, then using "bash -x"):

[… rocketchat has just been backed up and was restarted by the backup command …]
trolls:~ # rocketchatctl update
Updates could be risky, you can use the backup option # rocketchatctl backup, to create a backup of the rocketchat database first. Please check that you have enough space in your system to store the backup.
/usr/local/bin/rocketchatctl: line 900: [: 3.9.0: unary operator expected
Could not determine if updates available for RocketChat server.


[… immediate retry with debugging enabled …]
trolls:~ # bash -x rocketchatctl update
+ set -o nounset
+ main update
+ local VERSION=latest
+ local -r NODE_VERSION=v12.14.0
+ local -r NPM_VERSION=6.13.4
+ local MONGO_VERSION=
+ local -r 'MONGO_URL=mongodb://localhost:27017/rocketchat?replicaSet=rs01'
+ local -r 'MONGO_OPLOG_URL=mongodb://localhost:27017/local?replicaSet=rs01'
+ local ROCKETCHAT_DOWNLOAD_URL=https://releases.rocket.chat/latest/download
+ local -r NODE_DEB_DOWNLOAD_URL=https://deb.nodesource.com/setup_12.x
+ local -r NODE_RPM_DOWNLOAD_URL=https://rpm.nodesource.com/setup_12.x
+ local -r ROCKETCHAT_DIR=/opt/Rocket.Chat
+ local -r ROCKETCHAT_DIR_UPDATE=/opt/Rocket.Chat.bkup
+ local -r ROCKET_VERSION_INFO_URL=https://releases.rocket.chat/latest/info
+ local -r ROCKETCHATCTL_DOWNLOAD_URL=https://raw.githubusercontent.com/RocketChat/install.sh/master/rocketchatctl
+ local -r ROCKETCHATCTL_DIRECTORY=/usr/local/bin
+ local -r ROCKETCHATCTL_TEMP_DIRECTORY=/tmp
+ local -r TRAEFIK_DOWNLOAD_URL=https://github.com/containous/traefik/releases/download/v1.7.20/traefik
+ local -r CADDY_DOWNLOAD_URL=https://github.com/mholt/caddy/releases/download/v1.0.0/caddy_v1.0.0_linux_amd64.tar.gz
+ local distro=
+ local distro_version=
+ local install_node=0
+ local install_mongo=0
+ local rocket_healthy=0
+ local ROOT_URL=
+ local PORT=3000
+ local BIND_IP=0.0.0.0
+ local bind_loopback=true
+ local webserver=none
+ local install_node_arg=0
+ local use_mongo_arg=0
+ local webserver_arg=0
+ local rocket_mail_arg=0
+ local bind_loopback_arg=0
+ local rocket_domain=
+ local configure_rocketchat=0
+ local configure_webserver=0
+ local root_url_arg=0
++ date +%s
+ local sufix=1606910670
+ local node_path=/usr/bin/node
+ local backup_dir=/tmp
+ local update_node=0
+ '[' 0 -ne 0 ']'
+ '[' 1 -lt 1 ']'
+ [[ -t 0 ]]
+ case $1 in
+ os_supported
+ get_os_distro
+ '[' -r /etc/os-release ']'
++ . /etc/os-release
+++ PRETTY_NAME='Debian GNU/Linux 9 (stretch)'
+++ NAME='Debian GNU/Linux'
+++ VERSION_ID=9
+++ VERSION='9 (stretch)'
+++ VERSION_CODENAME=stretch
+++ ID=debian
+++ HOME_URL=https://www.debian.org/
+++ SUPPORT_URL=https://www.debian.org/support
+++ BUG_REPORT_URL=https://bugs.debian.org/
++ echo debian
+ distro=debian
++ . /etc/os-release
+++ PRETTY_NAME='Debian GNU/Linux 9 (stretch)'
+++ NAME='Debian GNU/Linux'
+++ VERSION_ID=9
+++ VERSION='9 (stretch)'
+++ VERSION_CODENAME=stretch
+++ ID=debian
+++ HOME_URL=https://www.debian.org/
+++ SUPPORT_URL=https://www.debian.org/support
+++ BUG_REPORT_URL=https://bugs.debian.org/
++ echo 9
+ distro_version=9
+ case "$distro" in
+ [[ 9 =~ ((9|10)) ]]
+ do_check_updates
+ get_rocketchat_latest_version
+ command_exists curl
+ command -v curl
++ curl https://releases.rocket.chat/latest/info
++ awk '-F"' '{print $2}'
++ awk -F: '{print $2}'
++ grep -w tag
+ latest_rocketchat_version=3.9.0
+ get_rocketchat_current_version
+ systemctl status rocketchat
++ awk -F= '{print $3}'
++ grep PORT
++ cat /lib/systemd/system/rocketchat.service
+ PORT=3000
++ curl --interface lo http://localhost:3000/api/info
++ cut '-d"' -f4
+ current_rocketchat_version=
+ print_make_backup
+ echo 'Updates could be risky, you can use the backup option # rocketchatctl backup, to create a backup of the rocketchat database first. Please check that you have enough space in your system to store the backup.'
Updates could be risky, you can use the backup option # rocketchatctl backup, to create a backup of the rocketchat database first. Please check that you have enough space in your system to store the backup.
+ [[ '' =~ ^([0-2]\..\..)$ ]]
+ version_gt 3.9.0
++ head -n 1
++ printf '%s\n' 3.9.0
++ sort -V
+ test 3.9.0 '!=' 3.9.0
+ '[' 3.9.0 == ']'
/usr/local/bin/rocketchatctl: line 900: [: 3.9.0: unary operator expected
+ print_check_updates_error_and_exit
+ errcho 'Could not determine if updates available for RocketChat server.'
+ echo 'Could not determine if updates available for RocketChat server.'
Could not determine if updates available for RocketChat server.
+ exit 1

I'm going to demonstrate the behavior with the following script, which is largely copy-and-pasted from "rocketchatctl", plus an "until" patch I'm suggesting (ignore the "--interface lo", that's a patch for another day):

trolls:~ # cat test_version.sh
#!/bin/bash

PORT=$(cat /lib/systemd/system/rocketchat.service |grep PORT |awk -F= '{print $3}')

current_rocketchat_version=""
until [ -n "$current_rocketchat_version" ]; do
                current_rocketchat_version=$(curl --interface lo http://localhost:$PORT/api/info 2>/dev/null |cut -d\" -f4)
                sleep 1
        done
trolls:~ #

When running this on our very much non-beefy stage system, it behaves as follows:

trolls:~ # bash -x test_version.sh ; systemctl restart rocketchat.service ; bash -x test_version.sh
[…can't reproduce the problem when rocketchat is still running…]
++ cat /lib/systemd/system/rocketchat.service
++ awk -F= '{print $3}'
++ grep PORT
+ PORT=3000
+ current_rocketchat_version=
+ '[' -n '' ']'
++ curl --interface lo http://localhost:3000/api/info
++ cut '-d"' -f4
+ current_rocketchat_version=3.9.0
+ sleep 1
+ '[' -n 3.9.0 ']'
[…restart happens here…]
++ cat /lib/systemd/system/rocketchat.service
++ awk -F= '{print $3}'
++ grep PORT
+ PORT=3000
+ current_rocketchat_version=
+ '[' -n '' ']'
++ curl --interface lo http://localhost:3000/api/info
++ cut '-d"' -f4
+ current_rocketchat_version=
+ sleep 1
+ '[' -n '' ']'
++ curl --interface lo http://localhost:3000/api/info
++ cut '-d"' -f4
+ current_rocketchat_version=
+ sleep 1
[…4 minutes 30 seconds later, when Rocket.Chat has finished starting up…]
+ '[' -n '' ']'
++ cut '-d"' -f4
++ curl --interface lo http://localhost:3000/api/info
+ current_rocketchat_version=3.9.0
+ sleep 1
+ '[' -n 3.9.0 ']'
trolls:~ #

So, I suggest a patch similar to the following:

--- /usr/local/bin/rocketchatctl        2020-12-01 08:14:38.872413023 +0000
+++ /usr/local/bin/rocketchatctl.2020-12-02.issue-34.patch      2020-12-02 16:52:39.787651510 +0000
@@ -757,7 +757,11 @@
 get_rocketchat_current_version(){
     if systemctl status rocketchat > /dev/null 2>&1; then
         PORT=$(cat /lib/systemd/system/rocketchat.service |grep PORT |awk -F= '{print $3}')
-        current_rocketchat_version=$(curl http://localhost:$PORT/api/info 2>/dev/null |cut -d\" -f4)
+        current_rocketchat_version=""
+        until [ -n "$current_rocketchat_version" ]; do
+                current_rocketchat_version=$(curl http://localhost:$PORT/api/info 2>/dev/null |cut -d\" -f4)
+                sleep 1
+        done
     else
         print_rocketchat_not_running_error_and_exit
     fi

I believe (but I'm not sure) we've also in the past experienced this problem during the post-update startup wait period.

[frdmn]

Hi @m4z. Thanks for reporting this. I have had come across this a couple of times in my local test environment as well. Will take a look at this ASAP.