[PATCH 13/13] Documentation: add a doc for blk-iolatency

From: Josef Bacik
Date: Tue May 29 2018 - 17:18:30 EST


From: Josef Bacik <jbacik@xxxxxx>

A basic documentation to describe the interface, statistics, and
behavior of io.latency.

Signed-off-by: Josef Bacik <jbacik@xxxxxx>
---
Documentation/blk-iolatency.txt | 80 +++++++++++++++++++++++++++++++++++++++++
1 file changed, 80 insertions(+)
create mode 100644 Documentation/blk-iolatency.txt

diff --git a/Documentation/blk-iolatency.txt b/Documentation/blk-iolatency.txt
new file mode 100644
index 000000000000..9dd86f4f64b6
--- /dev/null
+++ b/Documentation/blk-iolatency.txt
@@ -0,0 +1,80 @@
+Block IO Latency Controller
+
+Overview
+========
+
+This is a cgroup v2 controller for IO workload protection. You provide a group
+with a latency target, and if the average latency exceeds that target the
+controller will throttle any peers that have a lower latency target than the
+protected workload.
+
+Interface
+=========
+
+- io.latency. This takes a similar format as the other controllers
+
+ "MAJOR:MINOR target=<target time in microseconds"
+
+- io.stat. If the controller is enabled you will see extra stats in io.stat in
+ addition to the normal ones
+
+ - depth=<integer>. This is the current queue depth for the group.
+ - delay=<time in microseconds>. This is the current delay per task that
+ does IO in this group.
+ - use_delay=<integer>. This is how deep into the delay we currently
+ are, the larger this number is the longer it'll take us to get back to
+ queue depth > 1.
+ - total_lat_avg=<time in microseconds>. The running average IO latency
+ for this group. Running average is generally flawed, but will give an
+ admistrator a general idea of the overall latency they can expect for
+ their workload on the given disk.
+
+HOWTO
+=====
+
+The limits are only applied at the peer level in the heirarchy. This means that
+in the diagram below, only groups A, B, and C will influence eachother, and
+groups D and F will influence eachother. Group G will influence nobody.
+
+ [root]
+ / | \
+ A B C
+ / \ |
+ D F G
+
+
+So the ideal way to configure this is to set io.latency in groups A, B, and C.
+Generally you do not want to set a value lower than the latency your device
+supports. Experiment to find the value that works best for your workload, start
+at higher than the expected latency for your device and watch the total_lat_avg
+value in io.stat for your workload group to get an idea of the latency you see
+during normal operation. Use this value as a basis for your real setting,
+setting at 10-15% higher than the value in io.stat. Experimentation is key here
+because total_lat_avg is a running total, so is the "statistics" portion of
+"lies, damned lies, and statistics."
+
+How Throttling Works
+====================
+
+io.latency is work conserving, so as long as everybody is meeting their latency
+target the controller doesn't do anything. Once a group starts missing it's
+target it begins throttling any peer group that has a higher target than itself.
+This throttling takes 2 forms
+
+- Queue depth throttling. This is the number of outstanding IO's a group is
+ allowed to have. We will clamp down relatively quickly, starting at no limit
+ and going all the way down to 1 IO at a time.
+
+- Artificial delay induction. There are certain types of IO that cannot be
+ throttled without possibly adversely affecting higher priority groups. This
+ includes swapping and metadata IO. These types of IO are allowed to occur
+ normally, however they are "charged" to the originating group. If the
+ originating group is being throttled you will see the use_delay and delay
+ fields in io.stat increase. The delay value is how many microseconds that are
+ being added to any process that runs in this group. Because this number can
+ grow quite large if there is a lot of swapping or metadata IO occuring we
+ limit the individual delay events to 1 second at a time.
+
+Once the victimized group starts meeting it's latency target again it will start
+unthrottling any peer groups that were throttled previous. If the victimized
+group simply stops doing IO the global counter will unthrottle appropriately.
--
2.14.3