Time Machine backups slower in Sonoma than Monterey

The behavior you're seeing is not something we'd generally expect after upgrading from Monterey to Sonoma, but there are several factors that could contribute to the increased backup times.

• One of the most significant changes introduced in recent versions of macOS is the evolution of the APFS file system and Time Machine (TM) itself. Sonoma performs more extensive tracking of file system metadata, snapshots, and extended attributes than earlier versions. In some environments, this can result in Time Machine detecting substantially more changed items than were reported under Monterey, even when relatively little user data has actually changed.
• Another consideration is the age and size of the backup history. With approximately 2.7 TB of backup data spanning four years, TM has a very large backup database to maintain.
• Your storage hardware may also be contributing to the issue. Although the Seagate drives are perfectly suitable desktop drives, they are 5400 RPM disks, which offer relatively modest random I/O performance. TM performs a tremendous amount of small file and metadata operations rather than large sequential writes, so rotational latency becomes a significant bottleneck. Also, how the drive enclosure is connected to your Mac can be a limiting factor.

Regarding Disk Utility First Aid, a runtime of approximately 12 hours on a mirrored 2.7 TB volume made up of 5400 RPM drives is not necessarily abnormal. Since the second run completed without reporting additional issues, it suggests the file system itself is likely healthy.

As for whether upgrading to Sequoia would improve the situation, AFAIK, there isn't any evidence that Sequoia significantly changes TM performance in either direction. Here, at the ASC, some users report marginal improvements, while others see similar behavior to Sonoma. Based on your description, I would not expect Sequoia alone to reduce backup times from 8–12 hours back to the few hours you experienced with Monterey.

With all that said, here are some troubleshooting steps that may help:

• Verify the RAID volume's SMART health (if your enclosure exposes SMART data).
• Allow Spotlight indexing to fully complete after the OS upgrade.
• Check whether antivirus, cloud sync utilities, or other background applications are modifying large numbers of files. If your Mac is using a third-party AV app, disable, or better still, completely remove it.
• Review the TM logs to determine whether it's spending most of its time scanning, preparing, or copying files.
• If you're considering a hardware upgrade, replacing the 5400 RPM drives with SSDs (or at least 7200 RPM NAS-class drives) would likely provide a much more noticeable improvement than upgrading from Sonoma to Sequoia.

David Hollis2 wrote:

I'm not sure where to find the SMART data, if it's not in Disk Utility.

In your follow up reply you mentioned that both drives show "Verified" in the Disk Utility. That is what you want. A better method to get the full health statistics of those drives is to use a third-party app, like DriveDx. It's not expensive and it makes for a great tool to have in your "troubleshooting toolbox".

* The AppleScript routine doesn't log specific phases of a TM backup, but it is close to 50:50 copying and cleaning up. Maybe 60:40.

This is not where I was asking you to look. Instead, I am asking for the actual macOS logs, specifically around the TM process. There are a number of ways to do this, but I found the simplest was to use other third-party apps. Two that I suggest are provided free from The Eclectic Light Company. They are: TimeMachineMechanic & Mints. Both simplify the means to access the TM-specific logs so that you can get an idea on what the culprit(s) is(are). You can download them both from here: https://eclecticlight.co/consolation-t2m2-and-log-utilities/

Helpful References

• Time Machine Troubleshooting - Apple Community
• If Time Machine is slow on Mac - Apple Support
• If Time Machine takes a long time to back up - Apple Support

Regarding Disk Utility First Aid, thanks for the reassurance. It seems that Disk Utility can't check individual drives in a RAID array, only the combined array. It would be very useful to know the health of each individual drive.

This is where DriveDx will come into play, and again, I suggest that you get yourself a copy.

David Hollis2 wrote:

With DriveDx, I've done short and full self tests for both drives. All four completed without error. The full tests both took 8 hours.

Post the complete DriveDx text report for each drive in the RAID array and use the "Additional Text" icon which looks like a piece of paper on the forum editing toolbar. Usually with Hard Drives DriveDx will show a "Warning" or "Failure" notice for one of the SMART health attributes if there is a problem....I consider them as fatal because it is the beginning of the end for a Hard Drive. SSDs are a different matter.

The Raw Read Error Rate for one drive is 62% and the other is 68%.

Generally you cannot use this SMART attribute for assessing a Hard Drive's health because this attribute can differ between drives. If its "normalized" Value/Worst is at or below the Threshold (or at 001), then that would be a cause of concern, otherwise it is hard to use this attribute.

IMHO, it seems that the drives are OK. Sure, they could be better, 7200 vs 5400, but they're performing OK.

So, that begs the question about transfer speeds, and what could be slowing things down? The LaCie enclosure has a Thunderbolt 1 port. So there's a Thunderbolt 1/2 cable and an adapter to the Thunderbolt 4 port on the Mac. I think that the Mac acts as the RAID controller.

FYI, from the external RAID connection details shown in an earlier post here, the information shows a Link Speed of 6Gb/s, but the Negotiated Link Speed is just 3Gb/s. This means the connection is capable of about 500MB/s, but is only communicating at 250MB/s. This means those SATA Hard Drives are only connecting at SATA II speeds instead of SATA III speeds. This is a big problem.

Here is the information details from that earlier post:

Thunderbolt AHCI Controller:

  Vendor:	Thunderbolt
  Product:	AHCI Controller
  Link Speed:	6 Gigabit
  Negotiated Link Speed:	3 Gigabit
  Physical Interconnect:	SATA
  Description:	AHCI Version 1.19 Supported

Not sure if it happened in Sonoma, but I’ve seen where drives start connecting at slower USB speeds. Open up system information parentheses pull down option key and select it from the Apple menu), and look in the USB section for your backup drive. If it’s connecting at 480 Mbps, then for some reason it’s identifying incorrectly as a USB 2 device which will dramatically slow down writes. If it’s connected through a hub, I’ve seen entire hubs drop to USB 2 if a single USB 2 device is connected to the hub.

Ok, I did a preliminary review of the T2M2 results. Here are my findings/observations:

Destination IO performance measured: Wrote 1 50 MB file at 90.35 MB/s to "/Volumes/8 TB R1 External" in 0.553 seconds Concurrently wrote 500 4 KB files at 5.48 MB/s to "/Volumes/8 TB R1 External" in 0.374 seconds

One of the first things TM does, is perform a IO (input/output, aka, data transfer speed) test. It does this typically by writing a single 50 MB file to the backup destination drive. The transfer rate was recorded with an average transfer speed of 90.35 MB/s. For reference, I have a directly attached WD MyBook Pro. It has a transfer rate of 180 MB/s.

What is more concerning it the "Concurrently wrote 500 4 KB files at 5.48 MB/s" reading. My same WD enclosure was getting 875 MB/s for that same test.

There was nothing unusual with "Started 1 auto backup cycles, and 0 manual backups," or the "Created 1 new backups" and Thinned section. Same with snapshots. These are all normally expected.

Nor did I find anything unusual in the Backup Results section. The good news was that there were no errors reports.

Another performance-related test, still using T2M2, is to run it again, but this time make sure to select the "Speed" option. This will provide you with much more detail on the data transfer rates differ throughout the TM backup process.

Overall, I think what stands out is the concurrent write performance was somewhat dismal. TM definitely writes many small files during the backup process. Having low concurrent write performance can certainly slow down the overall backup. This is where I would start looking for an alternate solution for what you are using currently as your backup destination.

Also, as far as drive formatting for Macs, HDDs are best formatted in HFS+ (aka, macOS Extended (Journaled); SSDs are best formatted in APFS. This should be true whether you use drives for TM or general file storage, and should also be true with the newer versions of macOS. Unfortunately today, because of the demands for new data centers, things like storage devices and memory prices have increased dramatically ... so replacing them is a trade-off between lower performance vs. affordability.

Give me a chance to review your T2M2 results. In the meantime, I'm glad you had an opportunity to read through those references.

You will also find that the Eclectic Light Company has numerous articles, including those on using & troubleshooting TM. Two I would like for you to review, especially since they pertain to those tools you just used, are:

• https://eclecticlight.co/2024/01/10/time-machine-in-sonoma-detection-and-diagnosis-with-t2m2/
• https://eclecticlight.co/2020/07/17/time-machine-16-reading-a-normal-backup-in-catalina-using-mints/

They should unveil some of the mysteries behind the complexity of the TM process ... OR, make them more confusing! ;) What I would suggest is to first see what the "normal" backup process is, so you would have a baseline. Then, when issues occur, take another log snapshot to see what may have changed. Of course, at this point, you are already seeing an issue, so getting that first snapshot would be best after we can (hopefully) determine what is causing the current issue.