Added UI to System -> Boot which allows the user to tag a Boot Environment as "Keep" As keep tag is set it calls be = Update.FindClone("default"); Update.CloneSetAttr(be, keep = True) and handled accordingly.

This PR add the miimon polling rate during the creation of the bond interface. The missing miimon configuration prevent the failover of the interface.

This fix can be improved, the selection of monitoring and monitoring rate is now static, an addition to the UI can be useful to select the wanted monitoring type (ARP or MII), polling rate and/or arp target

This issue is referenced here in the forum

• Use AgentX instead of pass command to simplify code and improve performance
• Update the structure of MIB file to follow conventions for table
• Expose zvols state

Tickets: #24402, #25734

This is a draft PR, investigating if there are any side effect of setting zstandard as the default compression algorithm on pool creation (which is, by default, inherited by created datasets).

Notes:

• I'm not very sure about the need to manually set the featureflag in case of zstd

edit, after testing and checking things

• This shouldn't apply to the boot pool, as that one isn't created using the middleware afaik (middleware specifically lists these as being mounted under /mnt, which the bootpool isn't)
• Did some performance tests on an 2c/4t intel pentium CPU, 10gbe and 3x2disk mirrors. I don't think this is going to cause any significant slowdown on less-powerfull systems
• pools created from scratch on 12, aren't compatible with 12 anyhow in a lot of cases, so this wouldn't cause significant (new) compatibility issues (you can't import a pool created in TrueNAS 12/20.10 in 11 afaik)

Some actual data: General note: These numbers are supposed to be bottlenecked mostly by CPU and/or RAM. However they are not made to be a representation for performance, they are supposed to be used as a comparative reference only, meant to give a vague idea of performance delta. If you see weird results, it mostly is due to the test being CPU bottlenecked or test environment issues.

Allan Jude has done further research on performance here: https://docs.google.com/spreadsheets/d/1TvCAIDzFsjuLuea7124q-1UtMd0C9amTgnXm2yPtiUQ/edit#gid=0

His number are more based on a real-world hardware scenario. I highly suggest going through them.

Note about random performance: Random has been set to an 8K "extremely low" recordsize and mixed blocksizes.

Sequential Read:

Sequential Write

Sequential Read-Write

Random Read

Random Write

Random Read-Write

RAWR DATA

Compression Algorithm Performance.xlsx

test_results_1577651989.terse.txt

Test script: https://github.com/Ornias1993/zfs-compression-test

Note: For the IO slides, I've been lazy and column K is actually iops instead of bandwidth ;)

This PR adds changes to not allow encrypting ix-applications dataset even if parent is encrypted and also not allow migrating ix-applications dataset to another pool if source is encrypted as this results in various edge cases which after discussing with Caleb we have decided not to handle.

2 issues.

1. uuidd user/group don't exist so override the systemd unit file with nobody/nogroup since we don't need or use this service. If a user manually starts this service, it will at least start with the overridden user/group information.
2. mask uuidd.service and uuidd.socket because, again, we don't need or use it

Small purpose.

Two new alert services :

• ~AnyWebhookAlerteService : A way to call any arbitrary webhook to send alertes~
• FreeTelecomAlerteService : Alert service specifically designed to use the SMS api of the client of Free Telecom https://mobile.free.fr

A Pull Request introducing the alert servie exist into the WebUI.

Redmine: https://redmine.ixsystems.com/issues/31035

Much more detail in that redmine ticket but I will summarize here.

We are auto-mounting geli providers early in the boot process. While this isn't necessarily a problem on a single-node system, this causes problems for all TrueNAS HA appliances. Functionality for managing the destruction/mounting/decrypting of geli providers has been added to the webUI and middleware so I'm disabling this section in rc.conf.local

Add support for Hyper-V enlightenments which are important part of running Windows VMs on. I tried to follow TrueNAS philosophy of keeping it simple and hence the single flag to enable most commonly supported flags. Other option could be adding a option per each Hyper-V feature but that would quickly clutter the UI.

This is my first PR for TrueNAS so please instruct me if something is missing/incorrect.

This PR adds changes to use python to change user context in a process and then having a reusable implementation which can be used in other places as required to execute certain functions as some user.

Context

In docker we were not able to distinguish between a host path as required by the container itself to function or if it being a user specified host path.

With the change to containerd we can now make that distinction and paths have been updated accordingly so as to narrow down whitelisted paths which were allowed earlier.

Remove filesystem.file_get_contents since it doesn't have any type of validation and doesn't mix well with our rootless login functionality. This was a private method (only available across websocket connections) and was only used in our integration test suite, so I've fixed those tests accordingly.

It's not altogether clear how a user got in this situation, but the production impact of nuking the default domain idmap entry is rather profound. In the case of idmap-related tables we should remove the foreign key rather than deleting the table entry (at least if it's a builtin entry).

Two concerns:

• Exact partition sizes do not match with this approach (on 8GB HDD swap gets 900k smaller and data partition gets 900k larger), I guess it's alright?
• With incorrect arguments middleware will now SEGFAULT