Persistent queue corruption

[franck.lefebure]

Hi,
I had a disk saturation on a logstash 5.6.8 installation
I got to remove the "checkpoint.head" (0 bytes size) file to restart the pipeline
Now I have theses files in queue dir :

-rw-r--r-- 1 logstash logstash 262144000 Oct 10 10:07 page.3999
-rw-r--r-- 1 logstash logstash 262144000 Oct 10 10:53 page.4000
-rw-r--r-- 1 logstash logstash        34 Oct 10 10:53 checkpoint.4000
-rw-r--r-- 1 logstash logstash        34 Oct 10 11:39 checkpoint.3999
-rw-r--r-- 1 logstash logstash 262144000 Oct 10 11:39 page.4001
-rw-r--r-- 1 logstash logstash        34 Oct 11 16:19 checkpoint.head
-rw-r--r-- 1 logstash logstash 262144000 Oct 11 16:19 page.35

I would like to reinject in the pipeline the events in pages 3999/4000/4001.
I tried to repair directory with a pgrepair from a 7x install but without success

What are my options ?
(I actually struggle in Queue/Page/MmapPageIO sources to find a solution)

Thanks
Franck

[franck.lefebure]

Well..
I've debugged an enhanced the repair tool who can now rebuild a missing checkpoint.head with the informations from prévious pages
And I've ported it for a 5.6x
If you need it : logstash/PqRepair.java at flefebure-5.6 · flefebure/logstash · GitHub