This project is mirrored from https://github.com/firehol/netdata.git. Pull mirroring updated .
  1. 26 May, 2020 9 commits
  2. 25 May, 2020 8 commits
  3. 24 May, 2020 4 commits
  4. 23 May, 2020 6 commits
  5. 22 May, 2020 10 commits
  6. 21 May, 2020 3 commits
    • Adrien Mahieux's avatar
    • Jelmerrevers's avatar
      Update step-10.md (#9000) · 9de87052
      Jelmerrevers authored
      * Update step-10.md
      
      Kept on having an error on multiple Ubuntu server 20.04 instances, figured out it was due to a missing package in the installer.
      
      * Update docs/step-by-step/step-10.md
      
      * Add newline
      Co-authored-by: default avatarJoel Hans <joel.g.hans@gmail.com>
      9de87052
    • Austin S. Hemmelgarn's avatar
      Improve system-info.sh to better handle certain cases when gathering info on... · d98b2012
      Austin S. Hemmelgarn authored
      Improve system-info.sh to better handle certain cases when gathering info on the system's disk capacity. (#7902)
      
      * Remove trailing whitespace in system-info.sh.
      
      * Fix handling of APFS on macOS.
      
      APFS can have multiple volumes in a single partition, which means that
      the same functional 'volume' can appear multiple times in the output of
      `df`. Duplicate lines for such volumes will show the same total size and
      available space along with a common prefix for the device name.
      
      This updates the parsing logic for `df` on macOS to account for this by
      deduplicating lines in the `df` output that have the same total size,
      available space, and same normalized device name.
      
      This has the potential to incorrectly under-account space in some cases,
      but the liklihood of that happeing is much less than the certainty of
      overaccounting space on standard APFS configurations.
      
      * Properly handle VirtIO block devices when using /sys.
      
      The VirtIO Block device driver uses a dynamically allocated device major
      number, meaning that we can't trivially match on it.
      
      This updates the handling to properly look it up in `/proc/devices`
      instead of just using the whole dynamic device number range.
      
      * Add handling for NVMe block devices in sysfs code.
      
      They use dynamic major numbers just like VirtIO Block devices do.
      
      * Switch to device major discovery in /proc/devices for all device types.
      
      This converts the code to use `/proc/devices` to look up correct device
      major numbers for block devices taht we treat as disks just like we are
      already doing for those that have dynamically assigned numbers. This
      makes the code both more robust and easier to understand and modify.
      
      This also excludes some particularly old hardware that we were
      originally looking for. If needed, we can add in the required device
      names, but for now it's better to keep the list concise.
      
      * Correct handling of device major discovery.
      
      We need to strip leading whitespace before calling cut, not after.
      
      * Only use /sys/block if we can read /proc/devices.
      
      We use `/proc/devices` to do device number lookups that we then use to
      filter devices under `/sys/block`. As a result, if we can't read
      `/proc/devices`, then we won't actually parse anything out of
      `/sys/block` either, so we need to just fall back to parsing `df`
      output.
      
      * Deduplicate `df` output by device name on Linux.
      
      This ensures that we properly handle BTRFS subvolumes, counting each
      actual volume only once.
      
      * Use POSIX math expansion instead of awk to sum disk sizes.
      
      This avoids the rather annoying habit of AWK of printing integers in
      scientific notation instead of as exact values.
      
      * Correct `sed` options for POSIX complaince.
      
      * Fix disk info fetching for macOS>
      
      POSIX tools, as found on macOS, lack a number of rather useful filtering
      and sorting features, so we need to get rather creative with the
      handling on macOS to make the disk space computation work correctly.
      
      This unfortunately makes the calculation a bit less reliable than it
      would have been had the existing calculations worked correctly, but it's
      the best I can come up with without making things exponentiall more
      complicated.
      
      * Properly handle sector size when using sysfs.
      d98b2012