Between 9:00 PM and 11:59 PM Pacific time on Friday, September 8, 2017, the MySQL database software on each of our servers will be upgraded from version 5.6.36 to 5.6.37. This will cause an approximately 60 second interruption of service on each MySQL-using customer website at some point during this period.
This upgrade is necessary for security reasons and to fix bugs in MySQL. We apologize for the inconvenience this causes.
Update 10:21 PM Pacific time: The maintenance was completed as planned and all services are running normally.
Between 8:41 and 8:44 Pacific time tonight (August 29, 2017), there was a brief load spike on one of our servers that caused errors on some sites.
We’ve identified the cause of this (high NFS disk load caused by an internal tool we use) and fixed it so the problem won’t recur.
We apologize for the inconvenience this caused.
The authors of the Mailman mailing list software we provide for customers have recently released version 2.1.24 to fix several bugs.
We’ve upgraded the Mailman software on our servers as a result.
Users of Mailman lists shouldn’t notice any changes, but as always, don’t hesitate to contact us if you have any questions or see any problems.
We’ve added Let’s Encrypt SSL support to many Mailman mailing list pages, including all the pages that ask for passwords or login information.
The use of SSL should be transparent. If you have bookmarks to URLs on your own lists like http://lists.example.com/mailman/admin/listname, you’ll automatically be redirected to the SSL (https) version of that page when you load it.
(New lists may not use SSL immediately when created, but will be switched to SSL when our system is able to obtain an SSL certificate for them. Quicker automation of this is something we’re working to improve.)
Users of Mailman lists shouldn’t notice any changes beyond pages being marked as “secure” in web browsers, but as always, don’t hesitate to contact us if you have any questions or see any problems.
Between 11:00 PM and 11:59 PM Pacific time on Thursday, June 29, 2017, each of our hosting servers will be restarted. This will cause a brief interruption of service (less than 5 minutes) for each site at some point during this period.
Read the rest of this entry »
Last weekend, we removed the final spinning disk from our hosting servers. We now use only solid state drives (SSDs) in the RAID arrays of all our web, database, and mail servers. This applies both to new accounts and to every existing account (unlike many companies advertising SSD hosting, who only do it for new accounts).
Most customers won’t notice any change recently, because we’ve been working on this project for a long time. We finished converting all MySQL database storage arrays to SSDs around a year ago, and many of our hosting and e-mail servers have used only SSDs for several months. But at some point, all sites hosted with us have gotten noticeably faster.
Using SSDs instead of spinning disks makes a big difference for shared hosting. It removes almost all cases of “latency”, which is when a script, database or mail program has to first wait for another site’s data to be read or written. For hosting servers, the drive access pattern consists of lots and lots of small operations scattered around the drives, requiring high IOPS. On spinning mechanical disks, this means waiting for the disk head assembly to move to the right place for each new file, during which time it can’t do anything useful. With SSDs, that “seek time” is completely eliminated, making the drives hundreds of times faster at small reads and writes. Hard disk seeks were once the main cause of delays and load on our servers, but that’s been almost completely eliminated. WordPress sites run faster, database writes finish more quickly, and mail loads without any delays.
Large SSDs aren’t cheap — they currently cost about the same per GB as hard drives did in 2008, and we’re storing more than ten times as much data per customer as we did then — but the result is worth it. We’re proud to offer high performance hosting to everyone.
Between 10:00 PM and 11:59 PM Pacific time on Thursday, May 25 2017, two of our hosting servers will be restarted: “web06” and “web12”. (This page explains how to find which server a site is on.)
This will cause a brief interruption of service (less than 5 minutes) for each site on these two servers at some point during this 2 hour period. Sites on other servers will not be affected.
Read the rest of this entry »
We’re continuing the process of upgrading our MySQL database servers from the 5.5 series to the MySQL 5.6 series (specifically, from version 5.5.54 to 5.6.36).
The upgrade requires that each Web server be taken offline for approximately 60 seconds, causing brief scheduled “downtime”.
Over the next few days, we’ll be updating these web servers:
- Wednesday May 3, 10:00-11:59 PM Pacific: servers ending in digit “5” (completed)
- Thursday May 4, 10:00-11:59 PM Pacific: servers ending in digit “4” (completed)
- Friday May 5, 10:00-11:59 PM Pacific: servers ending in digits “2” and “3” (completed)
- Saturday May 6, 10:00-11:59 PM Pacific: servers ending in digits “0” and “1” (completed)
(Servers ending in digits 6-9 have already been updated.)
Read the rest of this entry »
We’re beginning the process of upgrading our MySQL database servers from the 5.5 series to the MySQL 5.6 series (specifically, from version 5.5.54 to 5.6.36).
The upgrade requires that each Web server be taken offline for approximately 60 seconds, causing brief scheduled “downtime”.
This coming weekend, we’ll be updating some (but not all) web servers:
- Saturday, April 29, 8-10 AM Pacific: servers ending in digits “9” and “8” (completed)
- Sunday, April 30, 8-10 AM Pacific: servers ending in digits “7” and “6” (completed)
(Servers ending in digits 0-5 will be updated the following weekend; we’ll post a separate announcement about that.)
Read the rest of this entry »
Between 11:06 and 11:08 AM Pacific time today (March 17, 2017), the “web13” server experienced high MySQL database load that led to some sites being unavailable.
The problem has now been resolved, and other servers were not affected. We apologize for the inconvenience this caused our customers.
