Skip to main content
mySites.guru
New this monthChecklistsCoreFilestmp Folder ScanDirectAdmin ImportcPanel Quick AddDatabase ChecksStaff Offboarding

Slower on purpose: the Update Queue runs one update at a time per server by default

Slower on purpose: the Update Queue runs one update at a time per server by default

Update, 26 October 2026: the queue is now always on

Every update on every mySites.guru account now goes through the Update Queue, and there is no switch to find. Backups queue too: choose Back up, then update in the confirm dialog and each site is backed up in its lane before its updates run.

Select every out-of-date site, click update, walk away. It is the whole point of managing sites from one dashboard, and for most people it works fine. Our guide to mass upgrading Joomla and WordPress sites from one dashboard covers the whole workflow, and the bulk update Joomla page has the overview.

Update Joomla, WordPress and Every Plugin Across All Your SitesRead the full post

Then you get an account with a thousand sites on a single shared server, all of them selected, and the updates go out at once. The server does not come through that. On core updates it falls over under the load. On extension updates you can break an individual site instead, because two updates race inside the same install and neither expected company.

The Update Queue fixes that by doing less at a time. By default it runs one update per server, and everything else waits its turn. Every update on every account goes through it, and it will make your big batches take longer. That is the feature, not a side effect.

What actually goes wrong when you update everything at once

There are two separate failures here, and only one of them is about the server.

Start with that one. A Joomla or WordPress core update is not a small job: download, extract, run the post-install work, hit the database. One of those is unremarkable. Several hundred at the same moment on one box is a load spike, and shared hosting is where the sites-per-server count is highest and the resource headroom is smallest. The server goes down and it takes every site on it with it, including the ones you were not touching.

The other failure is the individual site. Two extension updates running against the same install at the same time can collide: both writing files, both touching the database, neither aware of the other. You end up with one site broken by its own maintenance.

What the Update Queue does

Every update you trigger joins a queue for the server that site lives on. By default one update runs, the rest wait, and the next starts the moment the previous one reaches a conclusion. It covers Joomla core, WordPress core, Joomla extensions, WordPress plugins and themes, and backups, and it does not matter how you started the update: the dashboard, the API and the MCP tools all go through the same queue.

You are not left guessing what it is doing. There is a progress bar in the header while a batch is running, a page showing each server with its running job and everything queued behind it, and anything that has not started yet can be cancelled.

The mySites.guru header bar with the update queue pill on the right: a circular refresh icon, a progress bar a little over half full and the counter 7/12, meaning seven of twelve queued jobs have finished

The pill only appears while there is something to watch. It counts finished jobs against the size of the batch you started, as 7/12, and any failures show separately in red beside it.

Three states of the update queue pill side by side: 0/12 with an empty track just after a batch starts, 7/12 with the bar partly filled, and 11/12 with the bar nearly full and one job still running

The same-site collision is fixed too. One site never runs two updates at once, so a core update and an extension update on the same site cannot overlap, even on a server where you have raised the limit.

One lane per server, not one queue for everything

The queue is not a single global line. It is one lane per server, so a slow update on one host never holds up a completely unrelated host. Ten servers means ten updates running at once, one on each, until you raise the limit on any of them.

Hovering the pill opens the per-lane breakdown, so you can see which server is holding things up without leaving the page you are on.

The update queue popover under the header pill, which reads 7/12 and 1 failed. It lists three lanes: srv08.example.com with 1 running and 2 queued, web07.example.com with 1 running, 0 queued and 1 failed, and no known server (1) with 1 queued, above a View update queue link

That third row, no known server (1), is one of the shared lanes for sites that do not report a hostname. When the queue first shipped each such site had a lane of its own, and the reason that changed is below.

The detail worth explaining is what identifies a server, because getting it wrong would have been worse than not building the feature. Lanes are keyed on your account plus the server hostname, never the hostname on its own.

Hostnames are not unique. Hosting companies hand out the same generic names again and again, so the same string turns up across customers who have nothing to do with each other. In our own data one such name appears on 340 sites belonging to unrelated people, and another on 403. (Both are real counts from our platform; we have changed the names here.) Had we keyed lanes on the hostname alone, one customer’s slow update would have blocked a stranger’s, and neither would have been able to work out why.

About fifteen per cent of connected sites do not report a hostname at all. Those now share eight lanes per account. They started out with a lane each, per site, which stopped a site colliding with itself but put no ceiling on how many of them ran at once: an account with hundreds of such sites could fire hundreds of updates at the same moment, which is the exact thing the queue exists to prevent. Eight lanes keeps the ceiling, and it still never lets one unidentifiable site block anything else for long.

Raising the limit on a big server

One at a time is the right default for a shared server with hundreds of sites on it. On a dedicated server with twelve cores and a few dozen sites it leaves most of the box idle, so you can now raise the limit per server.

On the Web Servers page, click Edit on a server and set Parallel updates to anything from 1 to 8, or leave it on Default. The Updates: 1 at a time button in the page head sets that default for every server you have not set yourself. If most of your servers are big machines, raise the default and set the one shared host back to 1.

Save, and updates already waiting on that server start straight away rather than at the next pass of the queue. On the Update Queue page the server’s group shows a 3 at a time pill, and the header popover reads 2 of 3 running for that lane.

One site still never runs two updates at once, whatever you set. WordPress puts a site into maintenance mode for each plugin update, so a second update started on the same site while the first is running would fail, and Joomla’s installer is not safe to run twice at once on one site either. A higher limit lets more sites on that server update side by side, and each of them still gets one update at a time.

Sites we cannot match to a server stay in their eight shared lanes, one update at a time in each. We do not know which server they are on, so there is no server setting to raise.

Why is it on for everyone?

The queue shipped in August 2026 as a switch, off by default. The reasoning was that queueing is slower, and an account with a handful of sites spread across different hosts was never colliding in the first place, so turning it on for them would have meant waiting for a problem they did not have. It started to matter somewhere around twenty or more sites on the same server, which is where “update everything” and “take the server down” become the same action, and we left the call to the people who knew their own hosting.

Since 26 October 2026 every update on every account goes through the queue, and there is no switch. The lane design above is what makes that tolerable: lanes are keyed per account plus server, so the only updates that ever wait are yours, behind your own jobs on the same server, and a single update on an idle server still starts straight away.

No artificial delay

Queueing does not add artificial delay. A single update on an idle server is admitted immediately and starts within seconds, rather than waiting for the next tick of a timer. The waiting only appears when there is something ahead of you on that server.

Where is the queue?

Go straight to manage.mysites.guru/en/updates/queue, or press Cmd+K and use the hotkey u q. There is nothing to enable. The page shows your lanes, each with its running job and everything queued behind it, with filters for Running, Queued, Failed and Backups.

The mySites.guru Update Queue page grouped by server, working through 10 jobs: each lane shows its drain rate and time left, with one job Updating or Backing up and the rest Queued with their place in line, one Done and one Failed with Log and Retry buttons, under Updates and Cancel all queued buttons

The strip at the top sums up the run: working through 10 jobs, 1 done, 3 running, 5 queued on 3 servers, 1 failed, about 4 minutes left. Each lane shows its own rate and time left, and a WordPress core upgrade and an All-in-One WP Migration backup sit in the lanes alongside the extension updates, because they all go through the same queue.

Each row has a menu with Cancel, Retry, Log and Manage, and a Cancel all queued button at the top clears everything that has not started yet. The Updates button takes you back to the Updates page, where the header shows a Queue: N count while a batch is working through.

What it does not do

It does not retry failures

A failed update is marked failed, the error is kept and shown on the queue page, and the lane moves on to the next job. Nothing tries again on its own. An update that failed once usually fails the same way twice, and automatic retries on a half-finished update are how one broken site becomes several. Read the error, fix the cause, then use Retry in the row’s menu.

It does not reorder your work

Jobs run oldest first within their lane. There is no priority scheme to learn and nothing jumps the queue.

It does not make the updates themselves any faster or slower

Each update takes exactly as long as it always did. The only thing that changes is how many of them are allowed to happen at once.

It is not a substitute for backups

Backups still matter. Queued updates are still updates. Take a backup first, the same as you would if you were applying them one at a time by hand, which in effect is now what is happening.

The queue can do that for you. When a site you have selected has Akeeba Backup or All-in-One WP Migration connected, the confirm dialog on the Updates page offers a second button, Back up, then update. It puts one backup job per site into that site’s lane ahead of its updates, so the backup finishes before the first update starts. If the backup fails, that site’s updates are cancelled and stay on the list for you to deal with. Backups queue the same way and for the same reasons as updates: firing every backup on one server at once is the other reliable way to flatten it.

Seeing what actually happened

The queue tells you what is running now. The Update Log tells you what happened, and we have rebuilt it at the same time.

It lists the newest 2,000 updates applied across your sites, newest first, whether you triggered them yourself or mySites.guru applied them overnight. Each row shows the site, the extension and the version it went to, when it ran, and whether it finished or errored.

Filters for Failed, Pending, Overnight, By hand and Last 24 h narrow the list, so Failed pulls up every failure in one go. Check it the morning after a big batch. Updates you triggered do not always take, and this is where you find out which ones did. There is an Export CSV button if you need the list somewhere else.

It is at manage.mysites.guru/en/updates/scheduled. It has always been recording; it is just far easier to read now.

Who notices the difference

You will notice it if a meaningful number of your sites share a server, and especially if you have ever had a mass update coincide with a site or a server going down. On shared hosting or a reseller plan with a lot of sites on one box, this is aimed squarely at you: a big batch now takes as long as that server needs, one update at a time unless you raise it.

If you have a small number of sites, or they are spread thinly across many hosts, you will barely see it. Each server is its own lane, so sites on different hosts still update alongside each other as before, and a single update on an idle server starts within seconds.

If you are not sure which describes you, the Servers view groups your connected sites by the server they are on, which answers the question in about ten seconds. If one server has a long list under it, you are the person this was built for.

Thanks

A special thank you to Adi, who has been on my case about this one for a long time and was right to keep pushing. It turned out to touch a lot of layers in mySites.guru, from the dashboard and the API through to the background workers and the database lock that makes the per-server lanes safe, and that is why it took as long as it did to get right rather than only working.

Further Reading

Frequently Asked Questions

What is the Update Queue?
It is the lane every mySites.guru update goes through. Instead of firing updates all at once, mySites.guru runs one at a time per server by default: every update you trigger joins the queue for the server that site sits on, and the next one starts as soon as the previous finishes. It covers Joomla core, WordPress core, Joomla extensions, WordPress plugins and themes, and backups, and it applies to updates started from the dashboard, the API and the MCP tools alike.
Why is it on for everyone now?
Because the collisions it prevents happen on shared servers, and we cannot tell from the outside which of your sites share one. The queue shipped in August 2026 as a switch, off by default, so that an account with a handful of sites did not wait for a problem it did not have. Since 26 October 2026 every update on every account goes through it and there is no switch. It adds no artificial delay: a single update on an idle server still starts within seconds, and the only wait is behind your own jobs on the same server, which is exactly where the protection is needed.
Where is the Update Queue page?
Go to manage.mysites.guru/en/updates/queue, or press Cmd+K and type the hotkey u q. The page lists each lane with its running job and everything queued behind it, with filters for Running, Queued, Failed and Backups. Each row has a menu with Cancel, Retry, Log and Manage, and a Cancel all queued button clears everything that has not started yet.
Can I run more than one update at a time on a server?
Yes. On the Web Servers page, click Edit on a server and set Parallel updates to anything from 1 to 8, or leave it on Default. The Updates: N at a time button in the page head sets the default for every server you have not set yourself. Updates waiting on that server start as soon as you save. One site still never runs two updates at once, whatever the setting: WordPress puts a site into maintenance mode for each plugin update, so a second update on the same site would fail, and Joomla's installer is not safe to run twice at once on one site either. Sites we cannot match to a server stay at one at a time.
Does it retry an update that fails?
No, and that is on purpose. A failed update is marked failed, the error is shown on the queue page, and the queue moves straight on to the next job for that server. Nothing retries by itself. Automatic retries on a broken update tend to turn one problem into several, so retrying is left to you, from the Retry option in the row's menu, once you have read the error.
Will queueing make my updates take much longer?
A big batch takes as long as your servers need, and that is the trade you are making. The queue does not add artificial delay: a single update on an idle server starts within seconds rather than waiting for a timer, and each finished job admits the next one immediately. The wait only shows up when many sites share one server, which is the situation the queue exists for.
EU icon: AI MODIFIEDWritten and edited by a human, with AI assistance. Our approach to AI

What our users say

Pierre Cabrol
Pierre Cabrolmidiconcept.fr
★★★★★

The tools Phil has developed are so useful! It feels as though he anticipates everything we’ll need to manage our websites, whether they’re built on Joomla or WordPress. Well done! I’d recommend them without hesitation.

Read more reviews
Herzlich Nordisch
Herzlich Nordischherzlich-nordisch.de
★★★★★

As an agency this really helps to keep control, minimize work on several projects and act professional.

Read more reviews

Read all 309 reviews →

Get an email when the next post is published

Ready to Take Control?

One free audit of one site · no card · about 2 minutes to connect

Get Your Free Site Audit