On Sat, October 31, 2015 4:11 am, Matt Vander Werf wrote:
Would it be possible to also make it so that you can exclude parts of
lines/multiple lines (not just at the end of the message) from being
parsed
using a block form -like method, in addition to what you've already put in
place?
Something like this:
<!-- ncv_skip --> *this should be ignored* </--> ip6_mrt_cache : 0
ext4_inode_cache <!-- ncv_skip --> *this should also be ignored* </--> :
7960000
Hmm. Part of this is doable, but with the way it's structured I'm hesitant
to duplicate the searching logic once we've already progressed down to
identifying the label on a given line. This might have to run for every
single line of each status message that comes in, which means the cycles
do add up if we're not careful.
For the first case (where the tag as at the beginning of the line, it's
not too hard. Please try out the attached patch and let me know if it
works for you.
It adds a <!-- ncv_ignore --> processor which will move past the text
until it sees a "</-->" (or a newline), and will then continue on as
normal.
It also adds a block-level skip feature, which might be more efficient
(since we're just toggling a skip flag)
<!-- ncv_skipstart -->
these : 1
wont : 83
seen : 0
<!-- ncv_skipend -->
normal : 0
-jc
Basically, the parsing tool would just see this to parse for the graphing:
ip6_mrt_cache : 0
ext4_inode_cache : 7960000
Is this something that would be possible to also put in place? I'm not
sure
how the new functionality is done now, so I don't know if this is
something
that's feasible to add in or not.
If it's not very easy or very feasible or even possible at all (without
major re-work of the code), then that's fine. The existing new
functionality for parsing skipping will be very useful to me as it is!
But if it could be added, that would be really awesome and even more
useful
for me!!
If this is something that is possible, but probably won't be ready for the
actual 4.3.22 release, then that's perfectly fine with me as well! I'm
willing to wait until the next release if that's what is necessary! :)
Thanks as always!
--
Matt Vander Werf
On Sat, Oct 17, 2015 at 10:54 PM, J.C. Cleaver <user-87556346d4af@xymon.invalid>
wrote:
FYI, all,
4.3.22 will have comment markers that can be used for both skipping a
line
and for ending processing. Something along the lines of this will work:
<pre>
ip6_dst_cache : 7680
ndisc_cache : 7680
<!-- ncv_skip -->ip6_mrt_cache : 0
ext4_inode_cache : 7960000
<!-- ncv_end -->
scsi_sense_cache : 15360
</pre>
Not entirely sure if that will render properly :) But essentially:
pre-pend lines with "ncv_skip" in a regular HTML comment to skip it, and
a
bare "ncv_end" in a comment will abort further processing for that
message.
I'd considered adding regex capabilities directly into the NCV module,
but
it's a fairly time sensitive operation on more heavily-loaded servers,
so
it's important to keep existing modules like that as efficient as
possible.
The process of RRD module writing here is pretty simple though (simpler
than I'd recalled it), and it shouldn't be too hard to add other types
of
automated/regular processing besides NCV that could be called up.
HTH,
-jc
On Tue, October 6, 2015 7:42 am, Matt Vander Werf wrote:
(Sending this again to the list, as I sent it using the wrong e-mail
previously (non-subscriber e-mail))
I second this request for enhancement. It would be nice to be able to
send
more data from the client to be displayed on the column/test page, but
not
get parsed to be included in the RRD graph. Only parsing certain lines
that
are in the NCV format would be a nice feature. Something like what
J.C.
mentioned ("a data parser that looks for a hidden HTML comment marker
for
lines to process (and ignores the rest)") would work well for this,
IMO.
Or is there a way to do this already? (even through a hack?)
Nothing urgent, just something that would be nice to be included in
some
future Xymon release.
Thanks.
--
Matt Vander Werf
On Tue, Oct 6, 2015 at 10:39 AM, Matt Vander Werf <user-dfc3cf2ca434@xymon.invalid>
wrote:
I second this request for enhancement. It would be nice to be able to
send
more data from the client to be displayed on the column/test page,
but
not
get parsed to be included in the RRD graph. Only parsing certain
lines
that
are in the NCV format would be a nice feature. Something like what
J.C.
mentioned ("a data parser that looks for a hidden HTML comment marker
for
lines to process (and ignores the rest)") would work well for this,
IMO.
Or is there a way to do this already? (even through a hack?)
Nothing urgent, just something that would be nice to be included in
some
future Xymon release.
Thanks.
--
Matt Vander Werf
On Mon, Oct 5, 2015 at 7:48 PM, Boldt, David <user-945c7be1b8e4@xymon.invalid> wrote:
Thanks for these leads. Will looks at both the filter and "trends"
message options.
As a request for enhancement, I think it would be useful to have an
option to use a regexp to extract data from the status, one for each
variable.
On Mon, Oct 5, 2015 at 7:36 PM, J.C. Cleaver
<user-87556346d4af@xymon.invalid>
wrote:
On Mon, October 5, 2015 3:52 pm, Boldt, David wrote:
I'm in the process of converting Big Brother tests to XYMon and
would
like
to enable RRD/graphics for these tests.
Several of these tests present date/time information which is
important.
As a workaround for one test I have replaced the colon on a
HH:MM:DD
formatted date with a dot, but it doesn't really look like a time
anymore.
* Is there a way to have NCV ignore colons (and only use "=")?
* Is there a way to have NCV ignore a line (which might contain
colons)?
Unfortunately, there isn't a way to do either of these at the
moment.
I'd
considered a data parser that looks for a hidden HTML comment
marker
for
lines to process (and ignores the rest), but it's not quite present
yet.
* Can I use Data messages in such a way that the corresponding
Status
message is not parsed?
The easiest way to do this would be to add a --filter= option to
the
xymond_channel command line for the xymond_rrd process that's
handling
the
*status* channel, but not the data channel.
There's a CPU load hit for doing so (since now you're doing a PCRE
on
each
message), however it may not have much of an impact depending on
your
message volume.
* Might there be some other mechanism entirely?
Generally, speaking for more advanced RRD submission, the 'trends'
message
can be a useful data payload which is read directly by xymond_rrd.
The
advantage there (aside from more direct control) is the ability to
send
data points for multiple graphs at once, which could outweigh the
expense
of another message transmission.
HTH,
-jc
--
-- David Boldt
<user-945c7be1b8e4@xymon.invalid>
"Ah, this is obviously some strange usage of the word 'safe' that
I
wasn't previously aware of."
--Douglas Adams (1952 - 2001), Arthur Dent in "The Hitchhikers
Guide
to the Galaxy"