Unlock a world of possibilities! Login now and discover the exclusive benefits awaiting you.
Hi
After we upgraded to version 12.40 (SR3) we started to see alot of jobs failing with
Error: Connection aborted (System error: The handle is invalid.). We saw similar message with jobs using oledb connections aswell. Rerunning the job will make it run successfully
After a while these errors dissapeared (we could get like 9 failures every morning) but now they have come back. It is different jobs that fail and there is no pattern that i have seen to it.
I have had a case with Qlikview about it that didnt lead anyware and cannot find anything related to Qlikview when googling (there are articles about this error for Qlik sense though).
Anyone seen this and have some idea on what could cause these errors?
Regards Mats
@Mats_E This is most likely going to be environment related, either your disk/storage cannot keep up with the QDS service or other network latency related issues etc. If the common theme is things fail when more concurrent tasks are running, but they run fine later when there are few to no other tasks running, this is going to be the tell tale sign. The easy solution is to reduce the number of Engines for Distribution on the QDS service to keep things from overloading the environment, but the correct solution would be to read through the following link which will give you QDS best practices related to the QDS. I am guessing you did not find this in your searching, but it should give you a lot of things to check, I believe that should move you forward, not sure it will fix everything, but it should likely help significantly if you make changes based upon the recommendations in the whitepaper, it was written by a couple of the engineers, just FYI.
Shout if you have questions, the only other thing I could recommend would be to have Services come in and do a full health-check/use case review to help you figure out where things need to be. Not sure from what version you upgraded, but there have been some pretty huge changes in QDS service from 12.20 and later such that the infrastructure/environment has to be much more solid than it did back in the 11.20 track etc. Let me know if you have further questions, will try to answer as best I can, but I am pretty sure the most likely issue is going to be load on the environment cannot handle everything at this point with the higher I/O of the newer QDS services.
Regards,
Brett
Hi
I can give an update on this. The change that removed the issue was changing the configuration settings for how many times services communicated back to QVPR (ie read and updated the xml files). Dont remember which setting but it is mentioned in the Scaling pdf. During the different Qlikview versions the default (recommended value was changed) However when you actually upgrade it is not changed by the upgrade script so it will have to be changed manually. So we no longer had recommended value basically. After we change I wish that this info would be in upgrade Notes atleast.
Can recommend everyone have a read on on the scaling pdf as it has alot of good info on how to increase performance (and in some cases solve issues :))
Regards Mats
@Mats_E This is most likely going to be environment related, either your disk/storage cannot keep up with the QDS service or other network latency related issues etc. If the common theme is things fail when more concurrent tasks are running, but they run fine later when there are few to no other tasks running, this is going to be the tell tale sign. The easy solution is to reduce the number of Engines for Distribution on the QDS service to keep things from overloading the environment, but the correct solution would be to read through the following link which will give you QDS best practices related to the QDS. I am guessing you did not find this in your searching, but it should give you a lot of things to check, I believe that should move you forward, not sure it will fix everything, but it should likely help significantly if you make changes based upon the recommendations in the whitepaper, it was written by a couple of the engineers, just FYI.
Shout if you have questions, the only other thing I could recommend would be to have Services come in and do a full health-check/use case review to help you figure out where things need to be. Not sure from what version you upgraded, but there have been some pretty huge changes in QDS service from 12.20 and later such that the infrastructure/environment has to be much more solid than it did back in the 11.20 track etc. Let me know if you have further questions, will try to answer as best I can, but I am pretty sure the most likely issue is going to be load on the environment cannot handle everything at this point with the higher I/O of the newer QDS services.
Regards,
Brett
Hi Brett
I must have missed your reply. Sorry about that. Thx for the info. Ill check the link. And i think you are most likely correct in that it is due to performance. We have since the post decreased Qlikview engines from 20 to 10 and that has helped for a while but we are now seeing it more that last few days. We will check network latency etc aswell.
Do you have any more information about what changes was done to QDS after 12.20? When i compared the requirements prior to upgrading to 12.40 SR3 there were no changes in hardware requirements going from 12.20 to 12.40.
Regards Mats
Hi again
After having read that PDF file i must say i was impressed with the info in it. We have lots of things to look through and possibly change. Thx again
Regards Mats
Hi
I can give an update on this. The change that removed the issue was changing the configuration settings for how many times services communicated back to QVPR (ie read and updated the xml files). Dont remember which setting but it is mentioned in the Scaling pdf. During the different Qlikview versions the default (recommended value was changed) However when you actually upgrade it is not changed by the upgrade script so it will have to be changed manually. So we no longer had recommended value basically. After we change I wish that this info would be in upgrade Notes atleast.
Can recommend everyone have a read on on the scaling pdf as it has alot of good info on how to increase performance (and in some cases solve issues :))
Regards Mats