8
responses

Hello,

I'm currently facing an issue with missing log when log file rotate.
Same problem as described here:
https://nxlog.co/question/3495/missing-log-when-log-file-rotate
I'm using NXLog CE 2.10.2102. I tried to upgrade to last version (2.10.2150) but still same problem.
Could you tell me if this issue has been fixed in CE edition please? If not do you have any idea when it will be fixed?
Thank you very much.

Nicolas

AskedNovember 21, 2018 - 3:21pm

Answer (1)

Hello Nicolas,

NXLog CE 2.10.2150 is the latest version.
We never received word back from the other user in the linked issue as to the success of the new version in EE for their issue.
It could be useful to know if the EE trial works for what you are seeing as well.
It would also be useful to have some more information in this thread such as detail on environment and reproduction steps.

Thank you!

Comments (7)

  • Chewb's picture

    Hello,

    Thanks for your message.
    Actually I tried the trial EE version (4.1.4046) and it works fine. The bug has been fixed on EE version but not on CE.
    That's why i was wondering if you know when it will be reported on CE version.

  • Zhengshi's picture
    (NXLog)

    Understood.
    I verified yesterday that this is indeed on the roadmap, but I do not have an expected release date for it as of yet.
    We will post when we have additional information.

  • smoheiss's picture

    Hello,

    Im facing the exactly same issue than Nicolas. I have the last community version and tried the EE version. It works with the EE but not with the CE version.

    It seems that there are finaly several users facing this issue.

    May it be a legitimate reason to accelerate the fix of this issue?

    Thank you.

  • smoheiss's picture

    Hello,

    It seems that if the file monitored by nxlog got data between two "file check" and got rotated, then I lost data.

    I tried to reduce the PollInterval.

    Of course, if I put something like PollInterval 0.1, I reduce the probability of losing logs but I consume more resources.

    That's why I need the fix.

    I looked at the source code : i think the fix has to be implemented in the nx_file_input_close. But I'm not sure.

    My idea : in the im_file_check_file function in im_file.c, there is a call to the im_file_input_close when nxlog detect that the inode has changed. In this function, we can add a test : if the file size of the old file is greater than the file position known by nxlog, then the file must be read til EOF is found.

    Unfortunately, I donk know how to code that. But you are the experts :)

    Please can you fix it soon for Nicolas, for me too and all the users who encounter this issue. I'm pretty sure there are a lot of users who need this too but they are not aware of it.

    Moreover, can you give me the smallest pollinterval value we can specify?

  • b0ti's picture
    (NXLog)

    Thank you for pointing out the issue and the possible cause. We'll try to make sure it will be fixed with the next release. Please note that the NXLog Community Edition is a volunteer effort. If this is mission critical I suggest you consider the NXLog Enterprise Edition and our technical support services.