Snmp alerts for CPU utilization

Hi,
I want to enable snmp alerts on l2 and l3 switches to monitor CPU utilization.
I have Opmanager which is acting as SNMP server.
I have switches L2 and L3 which are running IOS 12.0, 12.1, 12.2
Do all these IOS versions support SNMP alerts?
And also I want to know the commands to be configured on switches for this.
Regards
skrao

You can configure SNMP traps for CPU Thresholding Notification.
http://www.cisco.com/en/US/products/ps6350/products_configuration_guide_chapter09186a0080455772.html
You should be ok with the versions you list but check exact IOS version supports it at http://www.cisco.com/go/fn
If you do not want to use traps then there are specific oids that can be polled for 1minute average (1.3.6.1.4.1.9.2.1.57) and also 5 minute average (1.3.6.1.4.1.9.2.1.58). I've used these in the past with no problems. These oids may have been superceeded so check for latest. You can always snmpwalk a device to check oids.
The cisco SNMP navigator is helpful when it comes to oids.
http://tools.cisco.com/Support/SNMP/do/BrowseOID.do?local=en
If you haven't got any SNMP configured yet on the switch check out http://forum.cisco.com/eforum/servlet/NetProf?page=netprof&forum=Network%20Infrastructure&topic=Network%20Management&CommCmd=MB%3Fcmd%3Ddisplay_location%26location%3D.1ddb4e54
HTH
--Phil

Similar Messages

Alerts for CPU,Memory Utilizations

Hi all,
In the last few weeks we have been encountering, Server Re-starts due to memory Shoots ups.
We are planning to have alerts for CPU, memory utlizations.
These would be triggered before a Standard Threshold is reached. This would avoid issues after threshold in production environment.
Are there any pointers, blogs etc available on how to achieve the same.
Thanks in advance,
Best regards
Abhishek

Basic alert funtionality of XI won't help. You must have CCMS configured in your landscape. It could be then used to raise alert for such issues. Have a look here
https://www.sdn.sap.com/irj/scn/go/portal/prtroot/docs/library/uuid/00c7270b-1111-2a10-198b-cf7adc86d695
Regards,
Prateek

SG-500 - SNMP OID for CPU Load?

Hello,
i have several SG500 (Standalone and stacked). How can i get the CPU Load via SNMP? Who knows the correct SNMP OID for CPU Load?
In the MIB which are provided as download a wasn't able to find the correct OID.

Hello,
There are three OIDs available for CPU usage:
CPU utilization for 5 seconds
.1.3.6.1.4.1.9.6.1.101.1.7.0
CPU utilization for 1 minutes
.1.3.6.1.4.1.9.6.1.101.1.8.0
CPU utilization for 5 minutes
.1.3.6.1.4.1.9.6.1.101.1.9.0
This came from a scrap note I had in an e-mail, so I am not exactly sure where this information is originally.
Let me know if that works for you,
Christopher Ebert - Network Support Engineer
Cisco Small Business Support Center
*please rate helpful posts*

SNMP OID for CPU and Memory Utilization on a MDS 9509

Does anyone know what the OIDs are for CPU and Memory utilization on a MDS 9509?
Thanks

CISCO-SYSTEM-EXT-MIB.my is a good place to start and you can determine the OID from the MIB.
Once you feel as though you are on the right track, have a look at:
http://www.oidview.com/mibs/9/CISCO-SYSTEM-EXT-MIB.html
I gather that what you need is:
1.3.6.1.4.1.9.9.305.1.1.1
and
1.3.6.1.4.1.9.9.305.1.1.2
Enjoy.
Stephen

SQL 2008R2 - Performance Condition alert for CPU Threshold event

I am trying to use SQLServer Agent alert for Resoure governor setup. I would like to be alerted when CPU threshold -> REQUEST_MAX_CPU_TIME_SEC > 20 -> is crossed for any particular query. I am in RTM version for SQL2008R2. For some reason
the alert never occurs thru SQL Agent Alert.
Here is my alert configuration
EXEC msdb.dbo.sp_add_alert
@name=N'CPU Alert',
@message_id = 0,
@severity = 0,
@enabled = 1,
@delay_between_responses = 60,
@include_event_description_in = 1,
@notification_message = N'Email from alert CPU',
@category_name = N'[Uncategorized]',
@performance_condition = N'MSSQL$XXXXXX:Workload Group Stats|Max request cpu time (ms)|default|>|1',
@job_id = N'00000000-0000-0000-0000-000000000000'
THe same alert is working if I put < 5, I am able to notified by email. But when I put > I dont see any alert.
I have tried changing the default setting for resource governor to various #s (1, 5, 20, 600) . But for any change, if i use perf condition > 5, i dont get any alert. If i use < 5 then i get alerts...
Is it a bug, or am i missing something...
Any help will be much appreciated...

I got the event notifications / service broker solution working. Thanks to
Aaron Bertrand blog for providing me with a working model.
I ended up using Aaron's code to create a more generic framework to support the processing of any number of trace events (not just exceeding resource governor's max cpu).
Here it is... use at your own risk:
-- Create the Target Service:
USE msdb;
go
CREATE QUEUE eventNotificationsQueue ;
GO
CREATE SERVICE eventNotificationsService
ON QUEUE eventNotificationsQueue ( [http://schemas.microsoft.com/SQL/Notifications/PostEventNotification] );
GO
CREATE ROUTE eventNotificationsRoute
WITH SERVICE_NAME = 'eventNotificationsService',
ADDRESS = 'LOCAL';
GO
-- Create the event notification:
-- SQL Trace Event Types: http://msdn.microsoft.com/en-us/library/ms186265.aspx
-- drop event NOTIFICATION [CPUThresholdExceededNotification] on server
CREATE EVENT NOTIFICATION CPU_THRESHOLD_EXCEEDEDNotification
ON SERVER
WITH FAN_IN
FOR CPU_THRESHOLD_EXCEEDED
TO SERVICE 'eventNotificationsService', 'current database';
GO
-- ****************** Add new notifications here ********************
--CREATE EVENT NOTIFICATION CPU_THRESHOLD_EXCEEDEDNotification
-- ON SERVER
-- WITH
FAN_IN
-- FOR XXXX
-- TO SERVICE 'eventNotificationsService', 'current database';
--GO
USE msdb;
GO
CREATE PROCEDURE [dbo].[dba_processEventNotifications]
WITH EXECUTE AS OWNER
AS
-- [dbo].[dba_processEventNotifications] - Provides a framework to alert on trace events.
-- Reference: Aaron Bertran's post at http://www.mssqltips.com/sqlservertip/2595/get-alerts-for-specific-sql-server-login-failed-events/
-- Revision History:
-- 20140827 D Turpin Original creation.
-- Candidate events include select * from sys.trace_events
BEGIN
SET NOCOUNT ON;
DECLARE
@message_body XML
, @message
NVARCHAR(max)
, @subject
NVARCHAR(255)
, @recipients
varchar(max);
--SET @recipients = '[email protected]; [email protected]';
SET @recipients = '[email protected]';
WHILE (1 = 1)
BEGIN
WAITFOR (
RECEIVE TOP(1) @message_body = message_body
FROM dbo.eventNotificationsQueue)
, TIMEOUT 1000;
IF (@@ROWCOUNT = 1)
BEGIN
-- Let's see if we can capture the query that is running:
EXECUTE sp_start_job 'DBA Load whoisactive';
IF (@message_body.value('(/EVENT_INSTANCE/EventType)[1]', 'nvarchar(max)') = N'CPU_THRESHOLD_EXCEEDED')
BEGIN
SELECT @subject =
N'A query on ' + CAST(@@SERVERNAME AS nvarchar(256)) +
N' exceeded the Resource Governor limit REQUEST_MAX_CPU_TIME_SEC';
SELECT @message = N'Event Notification: User query exceeded REQUEST_MAX_CPU_TIME_SEC at '
+ @message_body.value('(/EVENT_INSTANCE/PostTime)[1]', 'nvarchar(max)' )
+ CHAR(13) + CHAR(10)
+ @message_body.value('(/EVENT_INSTANCE/CPU)[1]', 'nvarchar(max)' ) + ' CPU (ms) '
+ CHAR(13) + CHAR(10)
+ N'SPID: ' + @message_body.value('(/EVENT_INSTANCE/OwnerID)[1]', 'nvarchar(max)' )
+ CHAR(13) + CHAR(10)
+ N' Check dba_baselines.dbo.whoisactive for currently active queries';
--Use this if you want to list the XML nodes/values.
--SELECT @message = CAST(@message_body AS NVARCHAR(max));
END
-- ********************* Add new events here ********************
--IF (@message_body.value('(/EVENT_INSTANCE/EventType)[1]', 'nvarchar(max)') = N'CPU_THRESHOLD_EXCEEDED')
--BEGIN
-- SELECT @subject = = N'A query on ' + CAST(@@SERVERNAME AS nvarchar(256)) +
-- N' exceeded the Resource Governor limit REQUEST_MAX_CPU_TIME_SEC';
-- SELECT @message = N'Event Notification: User query exceeded REQUEST_MAX_CPU_TIME_SEC at '
-- + @message_body.value('(/EVENT_INSTANCE/PostTime)[1]', 'nvarchar(max)' )
+ CHAR(13) + CHAR(10)
-- + @message_body.value('(/EVENT_INSTANCE/CPU)[1]', 'nvarchar(max)' ) + ' CPU (ms) '
+ CHAR(13) + CHAR(10)
-- + N'SPID: ' + @message_body.value('(/EVENT_INSTANCE/OwnerID)[1]', 'nvarchar(max)' )
+ CHAR(13) + CHAR(10)
-- + N' Check dba_baselines.dbo.whoisactive for currently active queries';
--END
EXECUTE msdb.dbo.sp_send_dbmail
@profile_name = 'DBMailProfile'
, @recipients = @recipients
, @subject =
@subject
, @body =
@message;
END
END
END
GO
ALTER QUEUE eventNotificationsQueue
WITH ACTIVATION (
STATUS = ON
, PROCEDURE_NAME = [dbo].[dba_processEventNotifications]
, MAX_QUEUE_READERS = 1
, EXECUTE AS OWNER );
GO

SNMP, Alerts for OutOfMemory Exceptions

We are using WLS6.1 on Tru64. Using SNMP, can we generate alerts for OutOfMemory
Exceptions that occur on any Managed servers in a cluster?

We are using WLS6.1 on Tru64. Using SNMP, can we generate alerts for OutOfMemory
Exceptions that occur on any Managed servers in a cluster?The SNMP MIB defines the following entries in the jvmRuntimeTable:
jvmRuntimeHeapFreeCurrent INTEGER,
jvmRuntimeHeapSizeCurrent INTEGER,
you can set CounterMonitor on these attributes through the Console and set thresholds
to send you alerts when the heap reaches a certain limit.
hope this helps,
Mihir

Email alerts for CPU and Memory status of CUCM

Hi All,
can anybody help me here?
Will RTMT trigger the daily email notification for CPU and Memory status of CUCM? If yes, kindly let me know the procedure.
Thanks in advance !!!!

Hi Hariharan,
In addition to Atul's link please refer the below mentioned link CPU and Memory usage
http://www.cisco.com/en/US/docs/voice_ip_comm/cucm/managed_services/cucm_health.html#wp1101115
Tx,
Hope this helps
Shalu

Underlying tables for CPU utilization report

Does anyone know what tables/views these are based on?
Grid Control gives 31 day view etc but i would like to customise it for a range of days and therefore need to query the base tables.
Thanks in advance

This is the basis for any metric query. You need to do a little work to figure out the target_name, metric, name, key_value (if there is one)
You may or maynot require the second joins to mt2 and d2. It depends on the metric if they have a second value or not.
SELECT
t.target_name,
trunc(d.ROLLUP_TIMESTAMP),
          d.key_value,
round(max(d.VALUE_AVERAGE)) "% Used",
          round(max(d2.value_average)/1024) "FileSys Size (G)",
round(max(d.value_average)/100*max(d2.value_average)/1024) "Used (G)"
FROM
MGMT_METRICS_1HOUR d2, MGMT_METRICS mt2, MGMT_METRICS_1HOUR d, MGMT_METRICS mt, MGMT_TARGETS t
WHERE
d.ROLLUP_TIMESTAMP >= (SYSDATE - 31)
AND d.metric_guid = mt.metric_guid
AND d.target_guid = t.target_guid
AND mt.metric_column != 'Status'
AND mt.metric_column is not null
AND mt.metric_type = 0
AND t.target_type = mt.target_type
AND t.type_meta_ver = mt.type_meta_ver
AND (t.category_prop_1 = mt.category_prop_1 OR mt.category_prop_1 = ' ')
AND (t.category_prop_2 = mt.category_prop_2 OR mt.category_prop_2 = ' ')
AND (t.category_prop_3 = mt.category_prop_3 OR mt.category_prop_3 = ' ')
AND (t.category_prop_4 = mt.category_prop_4 OR mt.category_prop_4 = ' ')
AND (t.category_prop_5 = mt.category_prop_5 OR mt.category_prop_5 = ' ')
and t.target_type = 'host'
and target_name like 'sometestserver'
and mt.metric_name = 'Filesystems'
and mt.metric_column in ('pctAvailable')
and d.key_value like '/db05/oradata/SID'
and d.rollup_timestamp = d2.rollup_timestamp
and mt.metric_name = mt2.metric_name
and mt2.metric_column = 'size'
and d2.key_value = d.key_value
and d2.metric_guid = mt2.metric_guid
and mt2.metric_type = mt.metric_type
and mt2.target_type = mt.target_type
group by
t.target_name,
trunc(d.ROLLUP_TIMESTAMP),
     d.key_value

No Server Name displyed in mail notification Alert generated by "total CPU Utilization Percentage is too high" Monitor

Hello Everyone,
I need your assisstance in one of the issues with Mail Notification Alert generated by SCOM.
Mail notification received as below:
Alert: Total CPU Utilization Percentage is too high
Severity: 2
ServerName:Microsoft Windows Server 2008 R2 Standard - The threshold for the Processor\% Processor Time\_Total performance counter has been exceeded. The values that exceeded the threshold are: 97.091921488444015% CPU and a processor queue length of 16.
Last modified by: System
The Alert was generated by "Total CPU Utilization Percentage"
But the problem with the above mail notification is it doesn't mentions about the affected server. So I would like to know how to tweak the monitors to provide the server name into it.
Thanks & Regards,
VROAN

Hi
You can add alert source to the email format in the scom server.
refer below link for parameters
http://blogs.technet.com/b/kevinholman/archive/2007/12/12/adding-custom-information-to-alert-descriptions-and-notifications.aspx
Regards
sridhar v

WAAS experiencing high CPU utilization

We are monitoring WAAS devices for CPU utilization in our Network.Currently we have set the threshold for 70%. At this threshold we see a large number of alerts from WAAS crossing 70% CPU utilization. I guess 70% CPU utilization is normal for WAAS. Am I right???

Hi Daniel,
Yes, during the peak hours sometimes the CPU shoots even upto 95% and then in next 5 min comes down to normal below 70%.
Since our threshold is 70%, we get a lot of inactionable(because it becomes normal in 5-10 miinutes only) alert.I was wondering if we could change the threshold.
One of the Cisco Tac Engineer replied to the above question :
That is true. It is normal to have a high CPU utilization in normal Operation as we mentioned earlier it runs on underlying Linux kernel and linux does not use Page file rather it uses Maximum RAM that it could .
Regards,
Mohit

RV 082 or 042 V3 - cpu utilization

Hi.
Is there a way to get via web interface or via snmp data regarding CPU utilization, idle time etc for RV042 / RV082 V3?
I am using RV082 with 96 vpn tunnels on each unit and would be nice to have such info.
In RV220W I saw that in new firmware (1.0.2.4) there is a DASHBOARD web page.
Best regards,
Catalin Burla

Mohamed,
Do you have a local DNS server at this site or does the router handle your DNS queries?
Jasbryan

MGW error logs for Cpu utlization

Dear All
We are using Cisco AS5400 as mgw, we often found during call peak hrs thsi mgw's showing error logs for cpu utlization. I have following questions in regards to thsi error.
1. why this error logs comes?
2. where & what we need to chcek? any hardware or IOS issue?
3. main and important thing whether during such error logs there is load on MGW and due to over cpu utlization call which attempt on MGW may failed? if yesy its failed with MGW end or failed from source end as abondan? or else if MGW disconnect then what is ISDN casue code.
we are using PGW in signalling or nailed mode.
Please help me with this
Regards
Amit--

Hi
Thanks to both of you
Here is log for Cpu utilization status during peak time as we see this error logs during peak time only
show proc cpu sorted | excl 0.00
CPU utilization for five seconds: 78%/36%; one minute: 73%; five minutes: 70
P
ID Runtime(ms)   Invoked      uSecs   5Sec   1Min   5Min TTY Process
216   646931668 841788064        768 10.40% 9.81% 9.32%   0 CCH323_CT
211   723978128-1124958350          0 5.48% 5.51% 5.36%   0 VOIP_RTCP
77   437021944-2113488726          0 4.91% 4.60% 4.53%   0 IP Input
63   368724284 712553243        517 3.93% 3.96% 3.69%   0 ISDN
207   226701532 790289944        286 3.52% 2.89% 2.68%   0 VTSP
61      262232    529665        495 2.03% 2.16% 1.04%   0 AFW_application_
206   319808168-1186847347          0 2.62% 2.27% 2.16%   0 DSMP
233       37460     57790        648 1.63% 1.10% 0.02%   0 AFW_application_
   3      293876    595003        493 1.63% 2.38% 2.06%   0 AFW_application_
205    70108356 455758882        153 1.06% 0.72% 0.74%   0 Resource Monitor
245   129219428 972383769        132 0.90% 0.93% 0.01%   0 ISDN L2 Process
208    43246508 321780241        134 0.57% 0.38% 0.37%   0 TSP
   4    35595820 258014603        137 0.57% 0.32% 0.32%   0 EDDRI_MAIN
37   162961592 35154923       4635 0.40% 0.53% 0.56%   0 Per-Second Jobs
200    33113344 140290609        236 0.32% 0.56% 0.53%   0 CC-API_VCM
198    35068732   7031070       4987 0.16% 0.13% 0.13%   0 Compute load avg
175    15677516 144379255        108 0.16% 0.08% 0.08%   0 AAA SEND STOP EV
33    16841100 32895103        511 0.16% 0.07% 0.06%   0 Net Background
92     6052176 398473472         15 0.08% 0.04% 0.02%   0 RUDP Main Proces
On this MGW we have full E1 means 16 E1s & total calls we observed during this is around 458 to 482 active callslls.
our MGW IOS is 12.4(22) & we are not using Radius AAA.
Mpneers,
Can you tell me i can see few logs like callid rejected. means at that time that call rejected within MGW. so such failed call failed as what Abondon or what is ISDN cause code?
Please help me with this
Amit--

SNMP MIB for VPN3000 Memory

Dear All,
I would like to monitor the VPN3000 performance through HP Open View. I could find the MIB for CPU utilization, but couldn`t find the MIB for Memory usage. Already browsed the MIB Locator on CCO but couldn`t find anything. Does VPN3000 not have a MIB for Memory usage ?? I thought monitoring the memory usage is a common task.
Would be appreciate for any information.
Best Regards,
Engel

Thanks for the link. But I meant that I already loaded all the VPN3K MIBS to the HP OpenView. And from the HP`s MIB Browser I couldn`t find one for monitoring VPN3K`s memory usage.

AHCI cpu utilization skyrockets

This issue is a bit new to me--have done RAID and IDE setups for decades, but thought I'd tinker with AHCI. Motherboard is MSI 970a-G46. Enabling and disabling AHCI with an established Win7x64 installation is not a problem for me.
Problem is that after enabling AHCI properly, cpu usage soars to 25%-30%+ with the Windows AHCI drivers, and jumps to as high as 40% with the latest AMD chipset drivers. OK--this is what HD Tach reports, anyway. IDE settings for the same drives measure 1-2% cpu utilization. According to HD Tach, too, the performance of AHCI & IDE is identical. Ergo: I see no advantage for my client system running in AHCI and will return to IDE.
Agree--disagree? Suggestions? Thanks.

Quote from: Panther57 on 30-June-12, 01:01:20
This is an interesting post... With my new build I was set Raid0 / IDE. I had an unhappy line in device manager and changed to AHCI. Then it downloaded the driver.
I have not seen a jump in CPU usage. But I also have not been watching it like a hawk. Hmmm
I am going to watch my AMD System Monitor for results. In a post of mine..earlier.. I was told, and did some tests of AHCI vs: IDE. I ran IDE on my other PC (listed below, HTPC) and am now AHCI on my main 990FXA-GD80. The difference between the 2 ways tested on my 790FX actually did show an advantage IDE, using Bench32.
Not a huge advantage.. but a little over AHCI. I don't know if the difference is really worth much inspection.
I am looking forward to the results you get WaltC
Thanks, Panther57...;) My "results" are really more of an opinion, but ...
Right now I'm not really sure what hard drive benchmark I should be using or trusting!...;) HD Tach's last release in 2004 is now confirmed on the company's site as the last version of the bench it will make--as it is, I have to set the compatibility tab for WinXP just to run the darn thing in Win7x64! But...I installed the free version of HD Tune (and the 15-day trial for the "Pro" version of the program, too), and the results are very similar--except that HD Tune seems to be measuring my burst speeds incorrectly: HD Tach consistently puts them north of 200mb/s; HD Tune, well south of 200mb/s. (A strike against HD Tune--the free version does not measure cpu dependency--grrr-r-r-r. You have to pay for the "Pro" version to see that particular number, or install the Pro trial version which reveals those numbers for 15 days.)
OK, between the two benchmarks, and after several tests, cpu utilization seems high *both* in IDE and in AHCI modes. Like you, it has been quite awhile since I actually *looked* at cpu utilization of any kind for hard drives. I guess I wasn't prepared to see how cpu dependent things have become again. Certainly, we are nowhere near the point of decades ago when cpu utilization approached 100% and our programs would literally freeze while loading from the IDE disk, until the load was finished. The "good old days," right? NOT, hardly...;) I suppose, though, that with multicore cpus being the rule these days instead of the exception, cpu dependency is just not as big a deal as it was in the "old days" when we dealt with single-core cpus exclusively and searching an IDE drive could literally stop the whole show.
Again, when running these read tests to see the degree of cpu utilization, I found that while the tests were all uniform and basically just repeats of each other, done a couple of dozen times, the results for cpu utilization in each test were *all over the map*--from 0% to 40% cpu dependency! And the same was true whether I was testing in IDE mode or AHCI mode. That was kind of surprising--and yet, it still leaves open the question of how accurate and reliable the two HD benchmarks that I used actually are. Besides that, I did find a direct correlation between the size of the files being moved/copied and the degree of cpu dependency--the smaller the files copied and moved the higher the cpu involvement--the larger the files, the lower the cpu overhead in copying and moving, etc. Much as we'd expect.
So after all was said and done--it does seem to me that AHCI is actually more of a performer than IDE, albeit not by much. I think maybe it demands a tad less cpu dependency, too, which is another mark in its favor. In one group of tests I ran on a single drive (I also tested a pair of Windows-spanned hard drives in RAID 0 (software RAID) in AHCI and in IDE mode, just for the heck of it...;)), I found the *average* read speed of the AHCI drive some ~15mb/s faster than the same drive tested in IDE. That was with HD Tune tests. But as I've stated, how reliable or accurate are these benchmarks? Heh...;) Anybody's guess, I suppose.
My take in general on the subject (for anyone interested) is that going to AHCI won't hurt if a person decides to go that route, but it also won't help that much, either. You definitely can easily and very quickly move from an installed Win7 IDE installation to an AHCI installation, no problem (some sources swear it can't be done without a reformat and a reinstall--just not true! They just haven't discovered how easy and simple it is to move from IDE to AHCI and back again.) Current cpu dependencies whether in AHCI or in IDE surprise me they seem so high. However, the last time I paid close attention to such numbers was back when I ran a single-core cpu, and back then cpu dependency numbers for a hard drive meant quite a lot. Today's cpus have both the raw computational power and the number of cores to take that particular concern and put it on its ear, with a large grain of salt!...;)
I have three drives total at current:
Boot drive:
ST332062 OAS Sata, boot drive
then,
(2) ST350041 8AS Satas, spanned in software RAID 0, making two ~500GB RAID 0 partitions.
Total disk space ~1.32 terabytes, all drives including RAID 0 partitions running in AHCI mode. (Software RAID is just as happy with IDE, btw.)
My Friday "project" is complete...:] Hope I haven't confused anyone more than myself...;)

Long running GC - High cpu% utilization

I have a tomcat process running on linux which generates a lot of garbage.
Throughout the day the machine will be at around 80% idle, and then suddenly jump to 20% idle cpu. It can stay at 20% idle for 1-4 hours, then goes back to 80% idle where it can stay for hours as well. The graph looks like a step-function.
This does not correlate to the usage graph; the usage graph is smooth, typical of a B->C webapp.
In conjunction with the 20% idle, i also see about 20% spent in "system" cpu, as well as an increase in "user" cpu consumption. When at 80% idle, system% is around 1%-2%.
I log GC stats, and when the cpu is at 80% idle, i see:
75324.001: [GC 75324.001: [ParNew: 65408K->0K(65472K), 0.0102260 secs] 472166K->407009K(1535936K), 0.0104630 secs]
75327.998: [GC 75327.998: [ParNew: 65408K->0K(65472K), 0.0118190 secs] 472417K->407212K(1535936K), 0.0120560 secs]
75333.342: [GC 75333.343: [ParNew: 65408K->0K(65472K), 0.0105900 secs] 472620K->407480K(1535936K), 0.0108260 secs]
About 10 msecs spent in GC every 4 secs, but when it gets to 20% idle i see:
30336.581: [GC 30336.581: [ParNew: 65408K->0K(65472K), 0.5277730 secs] 244204K->203369K(1535936K), 0.5280800 secs]
30338.403: [GC 30338.403: [ParNew: 65408K->0K(65472K), 0.4632960 secs] 268777K->224890K(1535936K), 0.4635790 secs]
30340.256: [GC 30340.256: [ParNew: 65408K->0K(65472K), 0.5050930 secs] 290298K->248727K(1535936K), 0.5053670 secs]
So now it's taking 500 msecs every 2 secs or so.
I'm using 1.5 with the following options:
-Xms1500m -Xmx1500m -XX:+UseConcMarkSweepGC -XX:NewSize=64m
So a 1.5G heap, with a 64M young generation. My inclination is to bump up the size of the young generation, but when i see these long 500 msecs young gen collection times, i get worried.
Can anybody explain the strange step-function behavior of the process over time that would be strictly related to GC?
Note that i see exactly the same pattern using the jrockit jvm (although the absolute numbers are different). I see the same step-function graph for cpu utilization vs a smooth graph for usage.
Thanks.

The fact that you see the same behavior with two completely different VM's makes me think it's not the VM's. -XX:+UseConcMarkSweepGC will cause slightly ParNew collections to take slightly longer, when it is running, but not anything like 50 times slower.
Trying using jconsole to see what going on during the idle periods and the busy periods. That is, get thread stack traces. If you are running JDK-1.6.0, you can just start up jconsole and attach to the running VM. If you are running JDK-1.5.0, I think you need to start the VM with an additional option to allow jconsole to attach to it.

Snmp alerts for CPU utilization

Similar Messages

Maybe you are looking for