| Age | Commit message (Collapse) | Author |
|---|
|
Where there isn't a copyright statement, add one to make it explicit.
Also add license identifiers as MIT if there isn't one.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Use python 'with' symantics to ensure that the /proc/pressure/cpu file
descriptor used in SystemStats init is closed. Previously, this would
lead to a single file descriptor being leaked. For example:
ResourceWarning: unclosed file <_io.BufferedReader name='/proc/pressure/cpu'>
Signed-off-by: Aryaman Gupta <aryaman.gupta@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The last time of sampling would be updated within the SystemStats class
but not re-recorded into the datastore, leading to multiple samples being
collected in the same second in the sample function of buildstats.py.
Fix this to collect and store only one sample per second within a
certain tolerance to deal with variation in the arrival time.
This fix elimates the spikiness of sampled data, in cases where the difference
between the current and the last sample is taken. Previously, since many
samples per second were recorded, certain types of data would result in a
very small elapsed time and hence a small numerical difference. For example,
the CPU usage from /proc/stat is a running total of usage and taking the
difference between data collected 0.1 seconds apart would result in usage
appearing lower than it actually was.
Signed-off-by: Aryaman Gupta <aryaman.gupta@windriver.com>
Signed-off-by: Randy MacLeod <randy.macleod@windriver.com>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The Linux pressure monitoring system helps determine when system resources
are being overutilized by measuring how contended the CPU, IO and memory are.
This information can be found under /proc/pressure/ which contains 3 files -
cpu, memory and io. In each of the files, the format is as follows:
some avg10=70.24 avg60=68.52 avg300=69.91 total=3559632828
full avg10=57.59 avg60=58.06 avg300=60.38 total=3300487258
The "some" state of a given resource represents when one or more tasks are delayed
on that resource whereas the "full" state represents when all the tasks are
delayed. Currently, we only collect data from the "some" state but the
"full" data can simply be appended to the log files if neccessary.
The "avg10", "avg60" and "avg300" fields represent the average percentage
of time runnable tasks were delayed in the last 10, 60 or 300 seconds
respectively. The "total" field represents the total time, in microseconds,
that some runnable task was delayed on a resource.
More information can be found at:
https://www.kernel.org/doc/html/latest/accounting/psi.html
and in the source code under kernel/sched/psi.c
This commit adds functionality to collect and log the "some" CPU, memory and IO
pressure. The "avg10", "avg60" and "avg300" fields are logged without change.
In place of the "total" field, the difference between the current "total" and
the previous sample's "total" is logged, allowing the measurement of pressure
in between each polling interval, as was done for /proc/stat data. The log files
are stored in:
<build_name>/tmp/buildstats/<build_time>/reduced_proc_pressure/{cpu,io,memory}.log
mirroring the directory structure of /proc/pressure. If the /proc/pressure
directory does not exist or the resource files can't be read/opened, the
reduced_proc_pressure directory is not created.
Signed-off-by: Aryaman Gupta <aryaman.gupta@windriver.com>
Signed-off-by: Randy MacLeod <randy.macleod@windriver.com>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This adds SPDX license headers in place of the wide assortment of things
currently in our script headers. We default to GPL-2.0-only except for the
oeqa code where it was clearly submitted and marked as MIT on the most part
or some scripts which had the "or later" GPL versioning.
The patch also drops other obsolete bits of file headers where they were
encoountered such as editor modelines, obsolete maintainer information or
the phrase "All rights reserved" which is now obsolete and not required in
copyright headers (in this case its actually confusing for licensing as all
rights were not reserved).
More work is needed for OE-Core but this takes care of the bulk of the scripts
and meta/lib directories.
The top level LICENSE files are tweaked to match the new structure and the
SPDX naming.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
getVar() now defaults to expanding by default, thus remove the True
option from getVar() calls with a regex search and replace.
Search made with the following regex: getVar ?\(( ?[^,()]*), True\)
Signed-off-by: Joshua Lock <joshua.g.lock@intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
Some virtualized environments like Linux-VServer do not have the
entries under /proc that the new system usage sampling expected,
leading to an exception when trying to open the files.
Now the presence of these files is checked once before enabling the
corresponding data collection. When a file is missing, the
corresponding log file is not written either and pybootchart will not
draw the chart that normally displays the data.
Errors while reading or writing of data samples is intentionally still
a fatal error, because that points towards a bigger problem that
should not be ignored.
Reported-by: Andreas Oberritter <obi@opendreambox.org>
Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
Pre-processing /proc data during the build considerably reduces the
amount of data written to disk: 176KB instead of 4.7MB for a 20
minuted build. Parsing also becomes faster.
The disk monitor log added another 16KB in that example build. The
overall buildstat was 20MB, so the overhead for monitoring system
utilization is small enough that it can be enabled by default.
Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
Hooks into the new monitordisk.py event and records the used space for
each volume. That is probably the only relevant value when it comes to
visualizing the build and recording more would only increase disk
usage.
Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
/proc/[diskstats|meminfo|stat] get sampled and written to the same
proc_<filename>.log files as during normal bootchat logging. This will
allow rendering the CPU, disk and memory usage charts.
Right now sampling happens once a second, triggered by the heartbeat
event.That produces quite a bit of data for long builds, which will be
addressed in a separate commit by storing the data in a more compact
form.
Signed-off-by: Patrick Ohly <patrick.ohly@intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
Richard Purdie
Aryaman Gupta
Joshua Lock
Patrick Ohly