It is extremely odd J.C. and thanks very much for your reply, has given me something to think about. I am not at the office now but before I left, after the copying over of the rrds files from Friday all looked ok, graphs were being generated properly, from xymon, bbwin from hosts and devmon. Then an hour later just as I was leaving, I saw a few checks having the issue again. It was slowly starting again. I had decided I had to do the *restore* of the rrds though just for peace of mind about the network intervention at the weekend not being anything to do with this whole issue (which would be very unlikely in the first place) so now I can be sure that is not the culprit.
To answer your question, it is all (bar none actually) types of RRD (conn, disk, memory, devmon etc). I did not see anything unusual in the status history around that time, but now that it has happened again today after the restore, I have some good time stamps to check through log files tomorrow. Perhaps not all hosts/checks will be affected by the time I arrive at the office tomorrow.
At the vm level I have not checked yet (handled by another team) but will do tomorrow. I did check the server from within RHEL and cpu/memory/disk seemed fine today and last few days.
I still think it's (our) Xymon that's having some difficulties somewhere although general host memory corruption is something I will look at.
Thanks again, will post more when I make some discoveries.
Steve
On Tue, Jul 7, 2015 at 6:02 PM, J.C. Cleaver <cleaver at terabithia.org> wrote:
On Tue, July 7, 2015 5:13 am, Steve B wrote:
Hi all,
This weekend, something happened with all our graphs. Every hosts' graphs are either corrupted or distorted and the history is unusable. I have checked all the usual places for graphs logging, rrd-data.log and rrd-status.log and other system log files but I am stumped as to where to start fixing this. We are looking at restoring rrds from previous snapshot which may or may not work but still would like to solve this mystery.
I have attached 2 screens but I do not know if these are viewable on the mailing list. It is hard to explain without but essentially there are huge numbers in our graphs such 3945789385793485793847593847593847593847593847593845793485739 and lots of '?' and there is no usable history, just a straight line along the base with one peak (or two) around the time this all happened (with a day or two out either way). If you try to zoom in, you get to a screen that just says 'zoom source image' and it's a black screen but if you hover your mouse over the screen you can find an area that is selectable and this shows a close up of the zoom area
rrdtool info example (for the same screenshot host test):
filename = "disk,C.rrd" rrd_version = "0003" step = 300 last_update = 1436270189 ds[pct].type = "GAUGE" ds[pct].minimal_heartbeat = 600 ds[pct].min = 0.0000000000e+00 ds[pct].max = 1.0000000000e+02 ds[pct].last_ds = "89" ds[pct].value = 7.9210000000e+03 ds[pct].unknown_sec = 0 ds[used].type = "GAUGE" ds[used].minimal_heartbeat = 600 ds[used].min = 0.0000000000e+00 ds[used].max = NaN ds[used].last_ds = "28436524" ds[used].value = 2.5308506360e+09 ds[used].unknown_sec = 0 rra[0].cf = "AVERAGE" rra[0].rows = 576 rra[0].pdp_per_row = 1 rra[0].xff = 5.0000000000e-01 rra[0].cdp_prep[0].value = NaN rra[0].cdp_prep[0].unknown_datapoints = 0 rra[0].cdp_prep[1].value = NaN rra[0].cdp_prep[1].unknown_datapoints = 0 rra[1].cf = "AVERAGE" rra[1].rows = 576 rra[1].pdp_per_row = 6 rra[1].xff = 5.0000000000e-01 rra[1].cdp_prep[0].value = 4.4500000000e+02 rra[1].cdp_prep[0].unknown_datapoints = 0 rra[1].cdp_prep[1].value = 1.4218146600e+08 rra[1].cdp_prep[1].unknown_datapoints = 0 rra[2].cf = "AVERAGE" rra[2].rows = 576 rra[2].pdp_per_row = 24 rra[2].xff = 5.0000000000e-01 rra[2].cdp_prep[0].value = 2.0470000000e+03 rra[2].cdp_prep[0].unknown_datapoints = 0 rra[2].cdp_prep[1].value = 6.5402986560e+08 rra[2].cdp_prep[1].unknown_datapoints = 0 rra[3].cf = "AVERAGE" rra[3].rows = 576 rra[3].pdp_per_row = 288 rra[3].xff = 5.0000000000e-01 rra[3].cdp_prep[0].value = 1.2727000000e+04 rra[3].cdp_prep[0].unknown_datapoints = 0 rra[3].cdp_prep[1].value = 4.0657944878e+09 rra[3].cdp_prep[1].unknown_datapoints = 0
This weekend we had a network intervention in that we moved some network connections in one of the 2 data centers but there was no downtime as we switched the network connectivity to the other data room. Our Xymon server is running on a virtual server (RHEL5) and the version we are using is 4.3.19.
All graphs were fine until this point. Any ideas?
This is quite odd.
There aren't too many things that could concertedly affect all RRD's like that within the code path. Is it the same type of RRD (eg, disk) for all hosts, or all RRDs for all hosts? Did you see anything unusual in the status history snapshots (if any) taken around this time?
If it happened to RRDs on both the 'data' and 'status' channels at once, that narrows down the possibilities even further. I'm assuming you've checked syslog for host level events for the VM, but did anything odd happen with the hypervisor around this time? General host memory corruption is about the only thing I can think of that might cause this -- haven't run into it before.
Regarding fixing the issue, restoring from backups might be the easiest option. If you want to save the surrounding data, your best bet might be to export/reimport the RRD to remove the "spike". I've used http://www.serveradminblog.com/2010/11/remove-spikes-from-rrd-graphs-howto/ in the past for doing this. It's easiest to script around the various types of RRD files, using a similar max setting for all "la" graphs, for example.
I seem to recall someone posting a script they had used for this in the past, but a search of the list archives hasn't revealed anything for me.
HTH,
-jc