Handling WordPress Cron failures using wp-cli

To understand WP cron, let me quote the following text from Cavalcade documentation

wp-cron is not actually a real task scheduler, and doesn’t actually operate like cron. Instead, it’s a pseudo-cron system, which is run as a loopback HTTP call when you access a page on the site (essentially, the page “forks” itself to run scheduled tasks). This is fine for high traffic single-sites, but lower traffic sites might not have their cron activated if the site isn’t viewed.

As you can see, even if a site has moderate traffic, if the site uses aggressive caching techniques, it is possible for real traffic to never hit PHP or WordPress to process a request. When wp-cron fails, a number of things can fail… a scheduled may not get published on time, a newsletter may not be sent as scheduled, an important WordPress security may not be applied, a backup schedule may have been missed, etc.

Cavalcade offers an excellent way to prevent such failures. It is more suitable for multi-site networks with hundreds of cron entries. For single sites, it is easier to handle a failure in wp-cron using wp-cli tool and a server cron.

Most common alternative is to disable internal WP cron and trigger it externally using wp-cron.php file and a server cron. If you are using this alternative method, then please remove cron entry and remove the corresponding entry in wp-config.php file, as wp-cli wouldn’t work if internal cron is disabled.

Let’s make sure we are ready to use the internal wp-cron by running the following command…

wp --path=/path/to/wordpress cron test

You’d see a success message… “Success: WP-Cron spawning is working as expected.“.

Most common error message is… “Error: The DISABLE_WP_CRON constant is set to true. WP-Cron spawning is disabled.” In this case, please disable the line in wp-config.php that disables wp-cron.

Now, it is time to set up a server cron that runs every minute like this…

* * * * * /path/to/wp-cli --path=/path/to/wordpress cron event run --due-now &> /dev/null

 

The 5 stars indicate a schedule to execute the command every minute. The command that follows executes cron events that are due now. Please make sure you provide full path to wp-cli and your WP site installation directory. I have tested the above in multiple sites on various servers. So far, it has been the most effective way to handle the failures in WP-Cron gracefully on single sites. If you run a multi-site network or a complex site with hundreds of crons, I’d highly recommend Cavalcade.

WordPress Migration Checklist

A successful WordPress site gets migrated to another host or another server within the same host, at least once in every few years for a number of reasons. Better operating system, to upgrade the PHP version, to handle larger traffic, etc. The first step is to test your migration on a staging server. Most issues (such as theme / plugins incompatibility) can be caught, if we test our migration using an existing (full) backup of the site. So, the following list applies only to live site migration. Here are the list of areas to check after the migration…

Backups

There are hundreds of plugins available to take a complete backup of your site. Each one is prone to errors. If you visit the official wp.org support forum for each of the backup plugins, you’d understand what I mean. The new host (or the server) may have some limitations that may not work well the current backup plugin. That’s why it is recommended to use an external backup solution that doesn’t rely on WordPress core, or its plugins. I have a dedicated Github project to take backups of WordPress sites using simple bash scripts.

Four things to consider while checking backups.

  1. Make sure local database backups are taken correctly.
  2. Make sure local files backups are taken correctly.
  3. Make sure offsite backups are working.
  4. Make sure the backups are verified.

No backup is useful, if it is not verified for authenticity of what it contains. That’s why it is important to have a workflow to test the backups, with automated tests.

Email

Unlike backup (that is basically a complex process), email is relatively an easy process. That’s why there are thousands of plugins available related to emails in wp.org repo. Because, people use email in a WordPress site for a number of reasons. To send a upgrade email to administrator, to get messages via contact form, to send newsletters, to get informed about the new user registration, to get notified about new e-commerce order, etc.

Even though, email is as old as internet itself, reaching inbox is a complex process. Most common issue with WordPress sites is that the email never leaves WordPress in most cases and would be simply dropped at WordPress itself. That’s why the most used contact form plugin still recommends an add-on plugin to keep an archive of all messages that come via the contact form.

If you use a SMTP service and a plugin to send emails via WordPress, then, in most cases, the emails would continue to work, after the migration. It is still recommended to test emails by submitting a test entry via the contact form.

If you depend on the server or host to send emails, then you may be in trouble, as certain host blog ports related to sending emails. For example, Google blocks port 25 on its compute platform.

Other items to check

There are some less-priority items to check after the migration. Here’s an incomplete list…

  • Cron: Check if cron hasn’t stopped and continues to run events as expected.
  • Cache: Check if your full-page caching plugin caches posts and the web server serves cached content if exists. Server response time is the most critical factor. You may check it at https://web.dev/ or in Google Search Console that provides historical data.
  • CDN: If your CDN depends on the IP of the server and if your server IP changes, then it may not be able to pull the content. You wouldn’t even aware of this, as CDN caches contents for an average of 24 hours and checks for newer content only once in a day or week. So, you’d understand the CDN related issues only if you create and publish a new post with images.

I will add other items as and when I come across or if it is important. Do you have a checklist whenever you migrate? If I missed anything, please feel free to add it via the comments section.

Happy hosting!

DigitalOcean Hosting Review

No affiliate links are inserted in this post. I did not have and do not have any monetary benefits in posting a positive review about DigitalOcean (or any other hosts in this blog), either. No host is perfect. So, I am posting only what I like about DigitalOcean. There may be some cons of using any host. If you search the internet, you may find them. I use Google Compute Engine for hosting this site, though.

Logo - DigitalOcean

Time to Restart the Whole Server

Whenever I build a server from scratch, I make sure that the automatic security updates are enabled. Most security updates are applied automatically. However, certain updates, particularly, the updates related to Linux kernel needs a restart of the entire server. This means a short downtime of the site/s hosted in that particular server. No one likes a downtime. At least, Google Search Bots do not.

The server restart and the resulting downtime is the primary reason for the popularity of containers such as Docker, LXC, LXD, etc. Of course, there are other benefits of using a container. Not everyone can afford to use containers, load balancers, etc. Maintaining them cost a lot more than maintaining a single server without the container eco-system. Most individual site owners and small businesses opt for single server system to lower the cost of running a server.

In shared hosts, the Linux kernel updates are applied on-the-fly, using one of the multiple technologies available, such as kpatch, livepatch (commercial), etc. Most of them are commercial services. The free services come with a big warning such as the following…

WARNING: Use with caution! Kernel crashes, spontaneous reboots, and data loss may occur!

So, most vendors (shared hosts) opt for paid commercial services to apply Linux kernel updates without a complete restart of the server. If you use a shared host, then no worries. This post is valid only for VPS servers and dedicated servers. The term VPS may also mean cloud servers with certain hosts. Some popular VPS providers are AWS (Amazon Web Services), Google Compute Engine (GCE), Linode, DigitalOcean, Vultr, etc.

The time to restart a server (VPS or dedicated server) depends on various factors. The number of running processes, the number of running programs, the choice of web server, the database server, etc. Every tiny thing matters. Even the hypervisor used by the host can have an impact on the time to restart the whole server.

Choosing a host is hard. There are lots of factors to consider. Price, the location of data centers, the Operating Systems provided, the bandwidth or the network speed, overall value for money, etc. Among these, you might also want to consider the time to restart a server when choosing a host. Because, some hosts take more than 5 minutes to restart a server, resulting in a downtime that may be a considerable amount for Google Search Bots.

In my experience working with various hosts, DigitalOcean restarts the servers in the shortest time possible, often less than 30 seconds for a server with a memory of up to 32GB. An idle server with no sites hosted on it, may restart much more quickly, like less than 15 seconds.

There are other advantages of choosing DigitalOcean. I will update this post with those, if it is worth mentioning.

Happy Hosting!

Looking Back and Looking Forward

Looking back at 2018

With respect to work…

I switched the workplace from Srivilliputhur to Madurai, from BSNL broadband to Airtel broadband, from ADSL to VDSL. I have had a lot of downtimes with BSNL last year. Fortunately, I had JioFi, as a backup, at that time (I was one of the early adapters of Jio 4g in general). However, as more users joined Jio 4g network, the overall speed reduced after each passing month. I used Airtel around 10 years ago (when Airtel had only ADSL), when I was in Madurai. Never had any issue with Airtel.

Continue reading “Looking Back and Looking Forward”

Brotli on Nginx

Brotli compression was introduced by Google, exactly 3 years ago. It is yet to come by default with the standard Nginx packages of OS vendors or the official Nginx repo. Google maintains its own repo for brotli support on Nginx web server. However, it’s been two years since we had the last commit. Fortunately, a Googler, named Eugene Kliuchnikov, continues to develop a fork of Google’s repo. There are even plan to merge both repos. Still, there isn’t an easy way to compile Nginx from source.

Continue reading “Brotli on Nginx”

I am hiring!

I’ve been a freelancer since 2008. So, it’s been a decade. I am currently looking for a freelance system admin with keen interests in hosting WordPress and PHP sites. Even though, the hosting market, especially the WordPress hosting market, has some big names in it, there are still plenty of problems to solve.

If you have interests in open source technologies and if you have contributed to open source, that’d be awesome. But, it isn’t mandatory. If you know your stuff, or if you can learn stuff, that’s more than enough. At the same time, communication is a big part of our job.

Continue reading “I am hiring!”