Re: Occasional delayed output of events

On Tuesday, January 26, 2021 6:53:31 AM EST Burn Alting wrote: > On Tue, 2021-01-26 at 11:29 +1100, Burn Alting wrote: > > On Mon, 2021-01-25 at 19:20 -0500, Steve Grubb wrote: > > > On Monday, January 25, 2021 7:11:45 PM EST Burn Alting wrote: > > > > On Mon, 2021-01-25 at 18:53 -0500, Steve Grubb wrote: > > > > > On Saturday, January 23, 2021 5:55:44 PM EST Burn Alting wrote: > > > > > > > > How is the following for a way forward.a. I will author a > > > > > > > > patch to the > > > > > > > > user space code to correctly parsethiscondition and submit it > > > > > > > > on the > > > > > > > > weekend. It will be via a newconfiguration item to > > > > > > > > auditd.conf just in > > > > > > > > case placing a fixedextended timeout (15-20 secs) affects > > > > > > > > memory usage > > > > > > > > for users of theauparse library. This solves the initial > > > > > > > > problem > > > > > > > > ofausearch/auparsefailing to parse generated audit.b. I am > > > > > > > > happy to > > > > > > > > instrument whateveris recommended on my hosts at home (vm's > > > > > > > > and bare > > > > > > > > metal) to providemore information, should we want to > > > > > > > > 'explain' the > > > > > > > > occurrence, givenIsee this every week or two and report back. > > > > > > > > > > > > > > Seems reasonable to me. > > > > > > > > > > > > I can implement the 'end_of_event_timeout' change either asi. a > > > > > > command > > > > > > line argument to ausearch/aureport (say --eoetmo secs) andanew > > > > > > pair of > > > > > > library functions within the auparse() stable > > > > > > (sayauparse_set_eoe_timeout() and auparse_get_eoe_timeout())orii. > > > > > > a > > > > > > configuration item in /etc/audit/auditd.conf, or > > > > > > > > > > > > Which is your preference? Mine is i. as this is a user space > > > > > > processingchange, not a demon change. > > > > > > > > > > To be honest, I'm not entirely sure what we're seeing. I run some > > > > > teststoday on my system. It's seeing issues also. I'd still like > > > > > to treat theroot cause of this. But we do need to change the > > > > > default. That I whatI'm trying to figure out. > > > > > Back to your question, I'm wondering if we should do both? A > > > > > changeabledefault in auditd.conf and an override on the command > > > > > line. > > > > > > > > So far, all items in /etc/audit/auditd.conf appear to only affect > > > > thedaemon. Is this the right location to start adding > > > > non-daemonconfiguration items? (I accept there is no other place). > > > > > > ausearch/report/auparse all read the auditd.conf to find the canonical > > > location for where the logs are supposed to be. So, they already read > > > this file. I'd rather keep it there than make yet another config. The > > > only drawback it that it might again confuse people that auditd really > > > doesn't do anything with the records but just some light processing. > > > > OK. I will put it in /etc/audit/auditd.conf > > One question with this solution. If the user does not have read permission > to /etc/audit/auditd.conf, then any change cannot take effect. The default > mode for this file is 640 to root, so a non-root user could never change > the timeout. Right, but since they cannot access the logs, it's not a problem in general. But if they so happen to have a local copy of logs, then the command line override should allow them to correct this. I am also reviewing things to see if a better default can be picked. > Should I also add - a command line argument to ausearch/aureport (say -- > eoetmo secs) and, - a pair of new auparse() functions - > auparse_set_eoe_timeout() and auparse_get_eoe_timeout() > so that non root users can make use of the new configuration item. Yes, that is what I meant by doing both. We have default in auditd.conf that works for everyone with direct audit access. We have a commandline option for overriding the auditd.conf value. Although, I don't know why we would want to get the eoe_timeout value? I can't imagine a use for it right now. As for ausearch/report, let's just make a long option --eoe-timeout -Steve > Also, do you want the default timeout to be 2 seconds or should I make it > higher. I'm likely to adjust it, but I'm still looking to see what is happening. Just go with the 2 second default for now.

Thanks, -Steve