New LiveCharts locks up my computer

[hubanks]

i am running Windows XP Pro with the latest updated version of Norton Internet Security running (no i dont have a virus!)
i have an always on high speed internet (cable) connection and my other PC also running on the network doesnt hang up.

i run live charts for a few hours on my computer just fine ... then it just stops responding and completely locks up my computer ... and I have to reboot the system to free it up

Please bring back the old version of Live Charts ... that worked fine for me for years

[dannol48]

Hi,

How much real memory do you have in your computer and are you running any other applications besides LiveCharts on the system that experienced the lockup? Do you have any other Add-ons that might cause a conflict, i.e. check IE7 Tools Manage Add-ons for a list. You can disable Add-ons, but that might not be the problem here.

I'm running WinXP Pro with 1GB and have other services running.

I haven't experienced any lockups, BUT I am watching Windows Task Manager for a possible MEMORY LEAK problem with the app. I was pretty sure the product was doing that earlier in its development, which is why I've layed-back waiting for a more stable code base (until this crisis started).

I'm also watching NETSTAT for an odd release and reacquire of TCP sessions and rotation of sockets (as many as 6). One appears to be persistent with Winsock state of ESTABLISHED using socket 2812. The other 4 (or more at times) show predominantely state of TIME-WAIT with very low occassional state change to ESTABLISHED using socket 80/HTTP. However, it also appears the ones using HTTP are periodically closing and new session and socket opening (new socket # at my end). When that happens (about every 30 minutes), I get a jump of 5K or more in real memory use to the system. That might be where the memory leak problem is occuring......but, I'd need to do more intensive looks to be sure.

Oh...and since the app appears to use 2812 and 80, I don't think its a firewall problem (unless I'm missing something).

When did this Alpha product ever reach a "STABLE" code base (or was considered such by development)? I'm just curous when a Beta period was started/declared and how long it lasted, as there doesn't seem to have been any IMO. I was actually just getting interested when the added chart feature was added (and thank you for recognizing the need), so that code base is relatively new. A beta needs months of intensive scrutiny before release to production. I'm lost, as this is barely an Alpha with these problems. That's an observation from a professional of over 35 years and not meant to offend, but this is a product that many depend upon for their livelyhood.

Edit: And, what level of Java are you running?

Edit2: So far today, no lost sessions. I did do a reboot earlier, so I could get a clean system for new checks. I'm still using COMP$ for heartbeat keepalive. I had 3 secondary chart outages total yesterday, so that suggests something may have changed overnight. I may switch that off to see if they made any changes last night, i.e. force the problem. Sorry if I'm using your post for my observations.

[hubanks]

I have 2 GB of RAM

I am only running Live Charts
Norton Internet Security (firewall, trojan horse, virus protection) And Skype (I'm not using Skype it's just running in the background)

In terms of Add ons

Sun Java Console is disabled

The following add ons on enabled
Adobe PDF
HP Printer options
Norton Stuff
Shockwave Flash
Skype
Windows Messenger

(pretty basic stuff)

[hubanks]

java version 6 update 5
build 1.6.0_05-b13

[dannol48]

Well, that's the same release of Java I'm running.

I just captured the sequence of events that causes the secondary chart failures.

It does appear that the TCP session with socket 2812 fails. That causes an immediate reconnect sequence of all sessions. Visibly, the secondary charts make an attempt to recover, but apparently do not recover. They drop the days data (in D period) and freeze. Toggling to another symbol and back does restore a failed chart.

In order to induce the failure, I simply had to switch the main chart to any symbol with a low tick rate. That caused failure in less than 2 minutes in one instance captured.

My best guess is a timer or the keepalive mismatch with Windows TCP, as I mentioned in yesterday's observations.

Using the DOW (2 second tick rate) or the Nasdaq (15 second tick rate) does seem to minimize the problem, which reinforces the concept that it's lack of traffic or timer from no data responsible.

Edit: In a sense there are TWO MAJOR BUGS....

1) The application apparently has a keep-alive problem unless induced/forced with a dataflow rate that perhaps mimicks the old apps clock keep-alive.

2) Session failure recovery should be reconnecting both main page and all functions and the secondary chart pages. Secondary chart pages are currently failing. This indicates the app can't operate in real world, where occasional traffic bottlenecks and outages do occur. Those might account for my lower failure rate, even with heartbeat activity.

[hubanks]

hey thanks for all the debugging and feedback ... do you work for esignal ...
if not, it's sad ... they could use someone with your motivation and quick response time.

i am not even sure anyone from esignal has looked at my post (or most of the recent posts on this board)

[j1900q]

locks up

I am using Vista, same thing. Just locks and cpu usage goes to 100% and I am done til I reboot.
thanks
Keith

[eSignal_ToddI]

For those that are experiencing this 100% CPU issue, is it just with IE 7.0?

How does Firefox or IE 6.0 run?

Java Console should be enabled in the add on manager.

[dannol48]

Memory Leak Problem

I did about 3 hours of testing yesterday (Sunday). I placed the main page and 4 secondary chart pages all on symbol CSCO. I used Windows XP Task Manager - Process Tab to observe the single iexplore.exe task (no other IE tasks running including shutdown of the esignal signon page). I also brought up a Command Prompt window and ran the following command: netstat 5 (which does an every 5 second list of TCP/IP sessions).

At start the task appears to use about 124MB. I monitored for about two hours and observed a steady creep in usage to around 160MB before I shutdown the test. The only activity I saw was the rotation of the socket sessions port number for the HTTP/80, as they were released and reestablished. Since it was Sunday, there were no other visible updates or dataflow changes that could account for the major increase.

The drop and reacquire of session is always a possilbe candidate for Memory Leak. It could be a coding problem or it could be this recent release of Java. It's just something for the programmers to investigate, as it might be the cause of the lockups and memory-depletion (heap) type messages from Java previously posted.

OH...and I started the Main + 4 at 7:37 this morning with 124,008K memory. At 9:22, it's already at 146,512K and steadily climbing.

Edit1: Corrected the K to MB on the Sunday observations. Sorry for the confusion.....too much coffee.

Edit2: 11:15 EDT At some point in the last 45 minutes or so, there was a sudden change in behavior. The Mem Usage of the app dropped to 122,764K and is somewhat rock steady state at that figure (which I've never seen in any prior observations). I have had several disconnects on secondary charts, and did switch from using Nasdaq to DOW (2 second heatbeat) in an attempt to keep charts stable and functioning. There were no other changes or refresh of the app or Windows at my end. I don't know if the switch to DOW tick was concurrent to the change in behavior on the leak. Trying to debug and do normal work at the same time is the eqivalent of the "rub your tummy and pat your head......ackward at best". If a change was made at the eSignal end, Thanks, and I'll try to keep reporting observations (if anyone there cares).

Edit3: 12:00 EDT Memory increased to 128,680K, but again appears to be more steady than creep (the 6M jump occurred just after my last observation report). All secondary charts and Main stable using DOW heartbeat. The socket session with port 2812 is stable and no change of port # at station end from prior report (probably due to DOW heartbeat).

Edit4: 14:00 EDT Two major changes in memory use occurred since last report. There was a sudden jump from 128MB to 146MB and a later change to 110MB. All secondary charts and Main remain stable using DOW heartbeat and no change to dynamic socket session port # for 2812 at the station end (very stable). One odd thing I just noticed was the DOW tick; it repeats 3 times for each tick, i.e. I'm seeing 3 ticks at exactly same time and value. Never seen that before.

Edit5: 14:30 EDT At approx. 14:10 trading/business system failed BSOD. Upon reboot and restart of all service apps, the strange behavior on Main with the repeats on the DOW ticker vanished.

Edit6: 15:25 Odd ticker problem with DOW has returned with 2 (not 3 per prior) ticks per report. All sessions also failed with only Main recovering. Socket 2812 session did reconnect, so it's consistent with old problems with an odd new DOW ticker situation.

[hubanks]

i enabled java console as an add on in MSIE

the problem remains however, the new Live Charts runs for a few hours ... then it just stops responding and locks up my computer

it appears the problem is in the Java application - or in some interaction between live charts and the java application. When i try to shut down, my computer says "Java Console not responding"

i have the latest version of java platform installed on my computer

Version 6, Update 5 build 1.6.0_05-b13)

[j1900q]

Locking up

I am using Firefox and it still locks up. 100% cpu usage. Please help.
thanks
keith

[hubanks]

so i can only get the new live charts to run for about 2 hours on my computer before everything locks up and i have to reboot.

i completely uninstalled java and reinstalled the latest version ... that didnt help

i running the latest version of windows xp pro with the latest copy of norton internet security

after i reboot live charts works fine for about 2 hours before locking up again

(tuesday april 22, 2008)

[hubanks]

i have to add that the old version of live charts would run for days without locking up my computer