OpenVMS Monthly Uptime Report

[Dante_5]

I have been looking for a solution to gather the monthly uptime for OpenVMS systems. I am thinking of a DCL procedure to keep track of this time. As we know when a system is rebooted, the OpenVMS Uptime is reseted and starts over.

I need to get a percentage for more than 100 hosts, in this is in regard of diverse SLAs we have.

I have found a couple of software tools that might help, Availability Mgr. from HP, and Heroix eQ Management Suite, but I think the use of these software seems to be too much just for getting this number, and they might not be worse buying.

Any helps or tips would really be appreciated.

Thanks.
Dante Villarreal

[John Gillings · Accepted solution]

Dante,

This depends on your definition of "uptime" and the granularity you require. You obviously know about F$GETSYI("BOOTTIME"), so it's easy to work out if the boot time is before the start of the current month (100%!), the tricky bit is if the boot time was within the month - how to work out how much time the system was up prior to the boot?

There are numerous places to look for time stamped evidence of a startup - OPERATOR.LOG, ACCOUNTNG.DAT, SECURITY.AUDIT$JOURNAL. You can either rely on things you expect to find on a "normal" system or put things there yourself. The biggest trick would be working out when an unexpected shutdown occurred.

The biggest problem would be finding matching pairs of time stamps that represent downtime.

Perhaps a simple mechanism which would measure time that the system was operating normally is a kind of "piggy bank". Give the system a token for each uptime sample.

Have a batch job execute at regular intervals (the interval will be the granularity of the measurement). Use a system logical name to keep track of the last execution, take the difference and write a record, with time stamp, to a file.
Your report can then be calculated for any time period by summing the records in your period of interest. If there is no previous time, then you must be the first execution since boot time.

There's also the "Uptimes Project" at http://uptimes.hostingwired.com/

I have no idea what their definition is, how the agent works, if you can access the information they collect directly from the node, or if it works without feeding the data back to the project web site.

[Dante_5]

Thank you so much for your help John. I am not really experienced in writing DCL procedures, but you certainly have given me good ideas and I am going to research on that.
Regards,
Dante

[Anton van Ruitenbeek]

Dante,

What we do is every day and during booting time we rename the operator.log to operator_._ . Because VMS is (in our case) creating at boottime this file, writing also when shutting down. When a system boots we need to check for more versions (also during the daily rename) . When you want to get track of the uptime (or downtime depends of how you looking) you can create a job to look simply across these filename in stead of digging these files.

AvR

[Stephen_67]

Hi Dante,

HP has a product for VMS called LAR (Local Availability Reporter) This reports uptime, if downtime was planned, unplanned etc. It produces monthly reports and emails them to you and also into eSMG (Electronic Site Management Guide) . If you contact your HP Services rep they will be arrnage for you to get a copy and see if it meets your needs

regards
Steve

[Mobeen_1]

Hello Dante,
If you consider uptime to be the time for which a node has been up and running. Then i would suggest that you use lexical function GETSYI and look for BOOTTIME on each of the nodes that you want to capture the uptime on and then you need to do some DCL.

Take current time in your DCL for each node and then subtract it from the boot time that you get from your lexical.

I would say that you start of on one node ... get the BOOTTIME and CURRENT TIME
Then subtract the BOOTTIME from CURRENT TIME. Once you deem that this is what you are looking for. Then you can look at other nodes/clusters.

If your clusters are spread across diff networks and time zones then you may run into other issues. Once you get there, throw the issue out here and i am sure, one of us will come up with a solution/idea/workaround.

All the best

regards
Mobeen

[Dante_5]

Many thanks to Stephen and Mobeen.
I had not heard of that product, i'll find out , sounds it might be the easy way.

Stephen: Do you know if this product is included with the Operating System at no extra cost ?

Mobeen: I am currently working on lexical functions. The most difficult part is to calculate the uptime for a calendar month, let's day, May 1st at 00:00 to May 31st 23:59. I need to calculate this, but considering that we have planeed downtime. If for example the host was up during the normal operational hours, and just except for the planned maintenance windows, then that host would have a 100% uptime.

[Stephen_67]

Hi Dante

It depends on the Level of Service you have with HP. In the UK we include it usages at varies levels within the cost of the Service Agreement, please feel free to ask you HP Service rep to contact me and I will provide them with the information.

Kind regards
Steve

[Wim Van den Wyngaert]

Also usable :

when you use performance advisor for the whole month, it will report the missing data timeframe, t.i. the time the system was unavaible (or data collector was down).

Wim

[Dante_5]

Thanks. Is this product "Performance Advisor" an HP product and have any link to it?

[labadie_1]

This product was a Dec product, has been sold to CA, and renamed Advise IT I think.

See

http://www3.ca.com/Solutions/Product.asp?ID=1174

regards


Géra

[Dante_5]

Thanks!

[Hein van den Heuvel]

>> The most difficult part is to calculate the uptime for a calendar month, let's day, May 1st at 00:00 to May 31st 23:59.

Correct. This has always been tricky.

But it got better in VMS V7.3-2 when a new lexical function F$DELTA_TIME was added.


Regular readers here will know I would now propose a 'perl' alternative, but it is a little tricky there also as there is no easy time-string to epoch-seconds function (or at least not that I can find today)
I would check out "vms_date_to_unix_epoch" in:
http://search.cpan.org/~dsugal/vms-misc-1_01/misc.pm

You also may want to look around in Openvms.org. It caries several interesting articles in the space for example:

http://dcl.openvms.org/stories.php?story=03/08/04/4042059


Hein.

[Dante_5]

Stephen:

Do you know if HP has another tool like "Local Availability Reporter", but one that can work on VAX ?
LAR only works in Alpha architecture. Maybe there could be an old tool from DEC.

If anybody of you know about this, would really be appreciated.
Thanks
Dante

[Martin P.J. Zinser]

Hello Dante,

instead of Advise IT you could run T4 on your nodes. Same principle, data for any downtime will be missing from your collected data. T4 is free, so there is quite a price differential to the CA product.

OTOH, if all you are interested in is "uptime" a simple script that updates an RMS record with the current time every minute or so would do the trick too. You would need to capture and save this record during system startup to determine the downtime.

Greetings, Martin

[Uwe Zessin]

That's a good idea, Martin. You could even write the boottime and the current time into an indexed file where the boottime is the primary key. You do a write the first time and then just issue updates against the same key. That way you even have a history in one file.

Hein, do you see a problem with repeated updates?

[Willem Grooters]

Uwe,

It's even better, for I learned there is an interface to get your applications write data to be included into T4 measurements. This may improve your reporting, where you can can add 'application uptime' to the reporting. What use is an up system if it cannot be used ;-)

Willem

[Hein van den Heuvel]

RMS has no problem with repeated updates.

The only very minor worry might be (not in this case) if you have a bunch of somewhat tightly packed records and they all grow during update. Maybe they are variable length and data fields get added, or maybe they start out blank and compression gets less effective as they get filled. For those cases you want some 'fill factor', or preload the record with garbage. In your case this might mean filling the 'update time' right way on the first put, but we are not talknig about thousands of systems to be monitorred are we? so no issue.

You can make the update very efficient by remembering the RFA after the put (or after a first find/get). However... you might want to consider a full open/find/update and maybe even a process creation for this job, as long as you don't do that every second, but every 5 minutes or so.
Why? Because this will prove the system was not only 'up' but was also able to perform a slightly complex tasks. If the system can do the update, but could not do the open (due to a disk lock perhaps) then would you consider the system to be up or down?

Finally... for these low usage applications consider a less effective, but easier to deal with key. Instead of the 'natural' VMS binary date key and time data fields, I'd use text in the YYYYMMDDHHMMSS format as that is readable and also sorts fine. I might even throw in some dashes or colon's to make is more readable still knowing full well this that this fixed fluff wastes recources :-).

Hein.

[Dante_5]

Where do I get T4 ? Does it run on both Alphas and Vaxes ?

I just got LAR (Local Availability Reporter) from HP, and it seems to be a nice tool so far. I still haven't checked the documentation completely, but unfortunatelly it just runs under Alpha.

Thanks to all!
Dante

[Martin P.J. Zinser]

Hello Dante,

T4 is at

http://h71000.www7.hp.com/openvms/products/t4/index.html

it looks like it is Alpha only (although I think it should be possible to do this on a VAX with a current VMS too, maybe ask hp if they have something that is not on the website)


Greetings, Martin

[Dante_5]

Thank you Martin,
I'll take a look at that. Primarily I am considering LAR (see above), but T4 is another option.
Please, for those of you who have worked with formerly Digital products, I would appreciate if there is a tool fo VAX platform.
Regards,
Dante

[Martin P.J. Zinser]

Hello Dante,

Big Brother allegedly has a VAX client. See

http://bb4.com/download.html

for details. Note: I do not have experience with this product.

Greetings, Martin