A Weird Imagination

Experimenting with ZFS

Posted in

The problem#

For my recent posts on ZFS, I wanted to quickly try out a bunch of variants of my proposed operations without worrying about accidentally modifying my real ZFS filesystems. Specifically, I wanted to know which ways of copying files would result in more efficiently reusing blocks from existing snapshots where possible.

The solution#

WARNING: The instructions below will modify the ZFS pool tank, which is the default name used in many ZFS examples, and therefore may be a real ZFS pool on your computer.

I strongly recommend doing all of this inside a VM to be sure you are not affecting any real filesystems. I used a VirtualBox VM that I installed Debian on and used the guest additions to share a directory between the VM and my actual machine.

First create a 1 GiB virtual (i.e. in a file instead of a physical device) ZFS pool to run tests on:

fallocate -l 1G /root/tank
zpool create tank /root/tank

Then perform various filesystem operations and inspect the result of zfs list -o space to determine if they were using more (or less) space than you expect. In order to make sure I was being consistent and make it easier to test out multiple variations, I wrote some scripts:

git clone https://git.aweirdimagination.net/perelman/zfs-test.git
cd zfs-test/bin
# dump logs from create-/copy-all- and-measure into ../logs/
./measure-all
# read ../logs/ and print space used as Markdown table
./logs-to-table --links
Create script orig rsync-ahvx rsync-ahvx-sparse rsync-inplace rsync-inplace-no-whole-file rsync-no-whole-file zfs-diff-move-then-rsync
empty 24K 24K✅ 24K✅ 24K✅ 24K✅ 24K✅ 24K✅
random-1M-file 1.03M 1.03M✅ 1.03M✅ 1.03M✅ 1.03M✅ 1.03M✅ 1.03M✅
zeros-1M-file 24K 1.03M❌ 24K✅ 1.03M❌ 1.03M❌ 1.03M❌ 1.03M❌
move-file 1.04M 2.04M❌ 2.04M❌ 2.04M❌ 2.04M❌ 2.04M❌ 1.04M✅
edit-part-of-file 1.16M 2.04M❌ 2.04M❌ 2.04M❌ 1.17M✅ 2.04M❌ 1.17M✅

The details#

Read more…

Splitting ZFS datasets

Posted in

The problem#

ZFS datasets are a powerful way to organize your filesystems. At first glance, datasets look a lot like filesystems, so you may default to just one or at most a handful per pool. But unlike with traditional filesystems where you have to decide how much of your disk space each one gets when it's created, ZFS datasets share the space available to the entire pool. Since datasets are the granularity at which ZFS operations like snapshots and zfs send/recv work, having more datasets can give you better control over having different backup policies for different subsets of your data, and ZFS scales just fine to hundreds or thousands of datasets, so you don't have to really worry about creating too many.

But if you're me (well, not just me) and you realize this after you already have months of snapshots of a few terabytes of data, how do you reorganize your ZFS pool into more datasets without either losing the snapshot history or ending up wasting a lot of disk space on redundant copies of data?

The solution#

Before doing anything with real data, make backups and confirm you can restore from them.

I do not have a one-size-fits-all solution here; instead I'll outline the general process and recommend you continually review at each step to make sure things look correct and be ready to zfs rollback and retry if you make a mistake or notice a way you could have done something in a more space-efficient manner.

  1. Create the new dataset hierarchy. I'll refer to the old dataset as tank/old and the new dataset root as tank/new.
  2. Do an initial copy of the earliest snapshot you want to keep from the .zfs directory. If it's @first, then the copy command will be rsync -avhxPHS /tank/old/.zfs/snapshot/first/ /tank/new/.
  3. Check your work and possibly delete or dedup files.
  4. zfs snapshot -r tank/new@first
  5. Do an incremental copy of the next snapshot. If it's @second, this may be as simple as rsync -avhxPHS@-1 --delete /tank/old/.zfs/snapshot/second/ /tank/new/, but that will waste space if you have moved files or modified small sections of large files.
  6. Check your work, and make any necessary changes.
  7. zfs snapshot -r tank/new@second
  8. Repeat steps 5-7 for each snapshot you want to keep.
  9. zfs rename tank/old tank/legacy && zfs rename tank/new tank/old

The details#

Read more…

Recreate moves from zfs diff

Posted in

The problem#

When doing an incremental backup, any moved file on the source filesystem usually results in recopying the file to the destination filesystem. For a large file this can both be slow and possibly waste space if the destination keeps around deleted files (e.g. ZFS holding on to old snapshots). If both sides are ZFS, then you can get zfs send/recv to handle all of the details efficiently. But if only the source filesystem is ZFS or the ZFS datasets are not at the same granularity on both sides, that doesn't apply.

zfs diff gives the information about file moves from a snapshot, but its output format is a little awkward for scripting.

The solution#

Download the script I wrote, zfs-diff-move.sh and run it like

zfs-diff-move.sh /path/ /tank/dataset/ tank/dataset@base @new

The following is an abbreviated version of it:

#!/bin/bash
zfs diff -H "$3" "$4" | grep '^R' | while read -r line
do
  get_path() {
    path="$(echo -e "$(echo "$line" | cut -d$'\t' "-f$3")")"
    echo "${path/#$2/$1}"
  }

  from="$(get_path "$1" "$2" 2)"
  to="$(get_path "$1" "$2" 3)"
  mkdir -vp -- "$(dirname "$to")"
  mv -vn -- "$from" "$to" || echo "Unable to move $from"
done

The details#

Read more…

Relative links in feeds

The problem#

In an RSS/Atom feed, relative links are a bad idea because it's unclear what they're relative to. There are ways to specify a base for them to be relative to, but since feed readers do not consistently respect those mechanisms, it's safer to just always use absolute URLs in feeds. And Pelican recommends setting RELATIVE_URLS = False to always generate absolute URLs. But that setting does not apply to the anchor links generated by the Markdown toc extension to link to headers.

The solution#

I wrote a Pelican plugin, absolute_anchors which rewrites all link destinations starting with # in every article to add the absolute URL of the article at the beginning of the link.

The details#

Read more…

Borderless browser window

The problem#

Web browser UIs have a lot more than just displaying the web page, which is useful when using them as a browser, but clutters the screen if all we want is to define what is displayed on part of the screen using HTML. So, can we get Firefox into a mode where it really does show just the website and nothing else? Firefox does have a fullscreen mode that does that, but it covers an entire monitor.

The solution#

To hide all of the Firefox menus and toolbars, put the following in the chrome/userChrome.css file under your Firefox profile directory (you will likely want to create a separate profile from the one you use for web browsing):

#TabsToolbar, #TabsToolbar-customization-target,
#nav-bar, #urlbar-container, #searchbar {
  visibility: collapse !important;
}

To hide the window border and titlebar, compile toggle-decorations.c and run

firefox &
./toggle-decorations $(xdotool selectwindow)

and then click on the Firefox window once it opens. It may be easier to bind it to a hotkey with xdotool getactivewindow or use some other way to identify the window.

The details#

Read more…

Fullscreen mode on part of screen

Posted in

The problem#

Many applications have a fullscreen mode that has a different interface from their windowed mode. For example, many media players will show just the video in fullscreen mode but include media controls in windowed mode. But, especially if you have a large monitor, you may want to use that interface while only having the application take up part of your monitor.

The solution#

I could not find a solution that works on every window manager.

The window manager handles resizing the application when it switches to fullscreen, so the most straightforward way to accomplish this is to not run a window manager. Problem solved! Unfortunately, window managers are really useful, so outside of some niche cases where you're positioning windows with xdotool, that's probably not what you want.

There's a "fakefullscreen" option in some forks of the very configurable window manager dwm: base dwm with the fakefullscreen patch always does fullscreen that way, the instantWM fork has a hotkey Super+Shift+F that toggles fake fullscreen for a window, and awesome can be configured to do the same.

For more common window managers, there is a solution, but more than two virtual monitors requires an xorg-server newer than 21.1.10 (which is the most recent release at time of writing, so you would have to compile it yourself), and in my tests, it only worked on Compiz, and not Mutter, KWin, or Xfwm. Use xrandr 1.5+ to define virtual monitors on sections of your monitors and then maximizing or fullscreening applications should respect those boundaries:

xrandr --setmonitor lefthalf 960/217x1080/132+0+0 LVDS-1
# This is a hack, should be "LVDS-1", not "none".
xrandr --setmonitor righthalf 960/217x1080/132+960+0 none

where the the geometery specification format is w/mmwxh/mmh+x+y (mmw/h="millimeters width/height") and LVDS-1 is the name xrandr gives to my physical monitor. Note that xorg-server 21.1.10 and older have a limit of one virtual monitor per physical monitor which we can circumvent by putting the second virtual monitor on "none".

The details#

Read more…

Status of long-running copy

The problem#

When running an incremental backup with rsync with the --progress flag, it often spends lot of time outputting nothing as it scans through many unchanged files. If you think of it before starting the transfer, --info=progress2 or the name2/skip2 --info flags would give more detail, but once the transfer has been going for a while, you probably don't want to cancel and restart it so you can add those flags.

The solution#

The documentation and this StackExchange answer say you can send a SIGVTALRM signal to rsync version 3.2.0+ and it will output its current progress, but that wasn't working for me.

As a workaround, you can use strace to get a running log of which files rsync is looking at, which includes files it skips without actually opening:

strace --attach="$(pidof rsync)" --trace=openat

(If that's not showing anything, try removing the --trace=openat filter and seeing if there's other syscalls with paths to filter on.)

Alternatively, this StackExchange answer suggests a way to see the currently open files including their sizes (including directories but not unchanged files being inspected):

watch lsof -p"$(pidof rsync | tr ' ' ',')"

(The same should work for a recursive cp/mv/rm.)

Similarly, for getting the status of a transfer of a single large file, this answer attempts to read the files cp is reading/writing to give a running percentage of how much it has copied; a similar approach might work for rsync.

The details#

Read more…

Hardlink identical directory trees

Posted in

The problem#

I will often make copies of important files onto multiple devices, and then later make backups of all of those devices onto the same drive. At which point, I now have multiple redundant copies of those files within my backup. Tools like rdfind, fdupes, and jdupes exist to deal with the general problem of searching a collection of files for duplicates efficiently, but none of them support only checking if files are identical if their filenames and/or paths match, so they end up doing a lot of extra work in this case.

The solution#

Download the script I wrote, hardlink-dups-by-name.sh and run it as follows:

hardlink-dups-by-name.sh a_backup/ another_backup/

Then all files like a_backup/some/path that are identical to the corresponding file another_backup/some/path will get hard-linked together so there will only be one copy of the data taking up space.

The details#

Read more…

Deleting deeply nested OPFS directories

Posted in

The problem#

The straightforward OPFS API for deleting a directory

await parentDir.removeEntry(name, {recursive: true});

doesn't work if the directory contains too many (several hundred) levels of nested directories. The sensible workaround is never create such a directory structure and wipe the OPFS storage entirely if you ever do by accident, but as discussed previously, I did get in that situation due to a bug and wrote some helpers to deal with it.

The solution#

For any reasonable real-world case, what you actually want is probably removeEntry():

await parentDir.removeEntry(name, {recursive: true});

or to delete everything from the root directory:

const root = await navigator.storage.getDirectory();
for await (const handle of root.values()) {
  await root.removeEntry(handle.name, {recursive: true});
}

or possibly to simply use your browser's settings to delete all site data, which should include OPFS data.

In case you do still want to delete a directory in OPFS without worrying about how deeply nested the directory structure is, you can use

async function removeDirectoryFast(dir) {
  const toDelete = [];
  let i = 0;
  let maxDepth = 0;
  for await (const fileHandle
             of getFilesNonRecursively(dir)) {
    maxDepth = Math.max(maxDepth, fileHandle.depth);
    toDelete.push(fileHandle);
  }
  async function deleteAtDepth(depth) {
    for (const f of toDelete) {
      if (f.depth === depth) {
        await f.parentDir.removeEntry(f.name,
                            {recursive: true});
      }
    }
  }
  const increment = 500; // Works empirically in Firefox.
  for (let depth = maxDepth; depth > 1; depth -= increment) {
    await deleteAtDepth(depth);
  }
  await deleteAtDepth(1);
}

This depends on the getFilesNonRecursively() helper from my previous blog post.

The details#

Read more…

Debugging OPFS

The problem#

While the web developer tools in Firefox and Chrome provide a Storage/Application tab for inspecting the local data stored by a web app, neither shows OPFS files there, making it difficult to tell what's going wrong when you have a bug (which was a problem when writting my recent blog posts about OPFS). There's open Firefox and Chromium bugs about the missing feature, so if it's been a while since this was posted when you're reading this, hopefully this is no longer a problem.

Additionally, the tools I did find all use recursion, resulting in them failing to work on the deeply nested directory tree I created by accident.

The solution#

If you don't have several hundred levels deep of nested directories, you can just use this Chrome extension or this script (or probably this web component, although I couldn't get it to install), all named "opfs-explorer".

The following AsyncIterator returns all of the files in OPFS without using recursion and adds properties to include their full path and parent directory:

async function* getFilesNonRecursively(dir) {
  const stack = [[dir, "", undefined, 0]];
  while (stack.length) {
    const [current, prefix, parentDir] = stack.pop();
    current.relativePath = prefix + current.name;
    current.parentDir = parentDir;
    current.depth = depth;
    yield current;

    if (current.kind === "directory") {
      for await (const handle of current.values()) {
        stack.push([handle,
                    prefix + current.name + "/",
                    current,
                    depth + 1]);
      }
    }
  }
}

And here's the simple HTML display function I've been using that calls that (you will likely want to modify this to your preferences):

async function displayOPFSFileList() {
  const existing = document.getElementById("opfs-file-list");
  const l = document.createElement('ol');
  l.id = "opfs-file-list";
  if (existing) existing.replaceWith(l);
  else document.body.appendChild(l);

  const root = await navigator.storage.getDirectory();
  for await (const fileHandle
             of getFilesNonRecursively(root)) {
    const i = document.createElement("li");
    i.innerText = fileHandle.kind + ": "
                  + (fileHandle.relativePath ?? "(root)");
    if (fileHandle.kind === "file") {
      const content = await fileHandle.getFile();
      const contentStr = content.type.length === 0
                      || content.type.startsWith("text/")
        ? ("\"" + (await content.slice(0, 100).text()).trim()
          + "\"")
        : content.type;
      i.innerText += ": (" + content.size + " bytes) "
                     + contentStr;
    }
    l.appendChild(i);
  }
}

The details#

Read more…