# of filesystems displayed in "disk" column is 1, but 5 in trends column
I feel like I've solved this problem before, but if so, I can't remember how. :)
I'm using some modified versions of the filerstats2bb perl scripts to collect data from my netapp filers, and ever since my upgrade from 4.1.2p1 -> 4.2.0, I no longer have 5 filesystems/graph displayed in the graph under the disk column. I do have 5/graph in the graph under the trends column, though.
Comparing the two URLs for the graph under the disk column, I see that "count=5" is included in 4.1.2p1, and "count=1" for 4.2. How is that value generated and passed to hobbitgraph.sh? Why is it different for the disk and trends columns?
Tom
On Mon, Aug 14, 2006 at 08:41:11AM -0400, Tom Georgoulias wrote:
I feel like I've solved this problem before, but if so, I can't remember how. :)
I'm using some modified versions of the filerstats2bb perl scripts to collect data from my netapp filers, and ever since my upgrade from 4.1.2p1 -> 4.2.0, I no longer have 5 filesystems/graph displayed in the graph under the disk column. I do have 5/graph in the graph under the trends column, though.
The "disk" column tries to count the number of entries in the disk status message, and decides how many graphs to include based on that. If you do a "view source" on the disk status webpage, you'll see a "<!-- linecount:N -->" string showing what count it arrived at.
The "trends" column looks at what disk-RRD files exist; if you have some filesystems that are not always reported, then there will be RRD-files for these, but they will be missing from the disk status message - hence the difference.
Regards, Henrik
Henrik Stoerner wrote:
On Mon, Aug 14, 2006 at 08:41:11AM -0400, Tom Georgoulias wrote:
I feel like I've solved this problem before, but if so, I can't remember how. :)
I'm using some modified versions of the filerstats2bb perl scripts to collect data from my netapp filers, and ever since my upgrade from 4.1.2p1 -> 4.2.0, I no longer have 5 filesystems/graph displayed in the graph under the disk column. I do have 5/graph in the graph under the trends column, though.
The "disk" column tries to count the number of entries in the disk status message, and decides how many graphs to include based on that. If you do a "view source" on the disk status webpage, you'll see a "<!-- linecount:N -->" string showing what count it arrived at.
Yup, I see a linecount of 1.
<!-- linecount=1 -->
The "trends" column looks at what disk-RRD files exist; if you have some filesystems that are not always reported, then there will be RRD-files for these, but they will be missing from the disk status message - hence the difference.
I wonder if it is caused by the names of the filesystems in the disk report. Seems like "case DT_NETAPP:" in do_disk.c is capable of handling them, although I'm not C code expert. Here's how the filesystems are listed in the disk report:
aggr0 aggr0/.snapshot aggr0total /vol/prod01/ /vol/prod01/.snapshot /vol/prod01/total /vol/vol0/ /vol/vol0/.snapshot /vol/vol0/total
On Mon, Aug 14, 2006 at 08:55:11AM -0400, Tom Georgoulias wrote:
I wonder if it is caused by the names of the filesystems in the disk report. Seems like "case DT_NETAPP:" in do_disk.c is capable of handling them, although I'm not C code expert.
That code handles it OK, that is why you get the 5 graphs on the "trends" page. The code that mis-handles it is the line-counting code that generates the "disk" status page.
Here's how the filesystems are listed in the disk report:
aggr0 aggr0/.snapshot aggr0total /vol/prod01/ /vol/prod01/.snapshot /vol/prod01/total /vol/vol0/ /vol/vol0/.snapshot /vol/vol0/total
Is there a red/yellow/green icon in front of those lines ? That would explain why you get a count of 1.
Regards, Henrik
Henrik Stoerner wrote:
On Mon, Aug 14, 2006 at 08:55:11AM -0400, Tom Georgoulias wrote:
I wonder if it is caused by the names of the filesystems in the disk report. Seems like "case DT_NETAPP:" in do_disk.c is capable of handling them, although I'm not C code expert.
That code handles it OK, that is why you get the 5 graphs on the "trends" page. The code that mis-handles it is the line-counting code that generates the "disk" status page.
Here's how the filesystems are listed in the disk report:
aggr0 aggr0/.snapshot aggr0total /vol/prod01/ /vol/prod01/.snapshot /vol/prod01/total /vol/vol0/ /vol/vol0/.snapshot /vol/vol0/total
Is there a red/yellow/green icon in front of those lines ? That would explain why you get a count of 1.
Yes, there is! Good catch! Should I drop those from the report? Or will that affect the column count in do_disk.c? Did something change from the 4.1 hobbit release that makes this more of an issue?
Tom
On Mon, Aug 14, 2006 at 09:10:07AM -0400, Tom Georgoulias wrote:
/vol/vol0/.snapshot /vol/vol0/total
Is there a red/yellow/green icon in front of those lines ? That would explain why you get a count of 1.
Yes, there is! Good catch! Should I drop those from the report? Or will that affect the column count in do_disk.c? Did something change from the 4.1 hobbit release that makes this more of an issue?
The reason is that in the Unix disk status, the lines with red/yellow icons are "extra" lines - they are at the top of the page to be seen, but the data is duplicated further down in the "df" output. So the line-counting code ignores lines that begin with an icon.
So the simplest solution would be if you could mimic that behaviour and have the iconised red/yellow lines first, and then the full report (including the red/yellow lines) after that.
Regards, Henrik
participants (2)
-
henrik＠hswn.dk
-
tomg＠mcclatchyinteractive.com