Hi, I'm currently using " listening_ports " table ...
# general
p
Hi, I'm currently using " listening_ports " table and " process_open_sockets " table with some joins with processes and related tables. The issue is that there's a hit and miss scenario where a connection might be open/closed between the query intervals and it may not represent actual data. To improve this, I'm thinking of using " socket_events " table instead with appropriate joins with process_events table etc. to replace both the queries/tables as stated above. Will it be sufficient or would I still be missing some data here ? Any suggestions are welcome. Thanks 🙂
y
Joining two '_events' tables might be problematic - for example a socket event occurred but for a process that already started (so it might not exist in process_events table). I think you should join socket_event with processes table. However, you might still miss some data if the process started and terminated in between the scheduled interval. Also note, the socket_events table is currently a linux only table, so if you use any other OS, this is something you should consider.
p
ok thanks, will try all these scenarios. We do have windows host as well, will check how can I improve data for win platform.
v
@yuvalapidot I think combining process_events, processes and socket_event should cover the scenario
do you see still see any data loss with above combination?
Also, if you are capturing the events and storing it in some database, then I think events gathered from process_events and socket_events will save you from data loss. Key part is capturing the events and storing some place safe (i.e database etc)
y
@Vibhor If the events are stored and you manage them, it is definitely possible to work with both events table, but if I understand correctly, it means you will have to implement the join yourself. As I see it, joining two events table in one query and scheduling it in osqueryd directly will not be sufficient. If I understand it wrong, please explain where is my misunderstanding.