I sometimes have an issue where my cluster freezes and items just stop processing. I am aware of what the bug is and waiting for the new release to account for it (Documented bug)
Today I saw something a bit weird… I know the exact time where things started to back up on the system About 30 minutes later I had to crash the node to get things to fail over quicker. When I reviewed the issue about 30 minutes prior (around the actual time where things started to go south) these files were created in my data dir. What are they? How do I read them?
#:/app/mysql/data # ls -la | grep 20
drwxr-xr-x 45 mysql mysql 4096 Jan 20 08:57 .
-rw-r----- 1 mysql mysql 67108864 Jan 20 08:32 0xaa5aafb1_data.000000
-rw-r----- 1 mysql mysql 67108864 Jan 20 08:33 0xaa5abdb3_data.000000
-rw-r----- 1 mysql mysql 67108864 Jan 20 08:33 0xaa5ae87c_data.000000
-rw-r----- 1 mysql mysql 67108864 Jan 20 08:34 0xaa5af6a4_data.000000
#:/app/mysql/data # file 0xaa5af6a4_data.000000
0xaa5af6a4_data.000000: data
1 Like
Likely you see gcache files
You do not need to read them, as they are only for internal usage
1 Like
@vadimtk
Thank you… I was unable to find via google what these files were. I wonder if this might be related to my intermittent crashes as according to PMM all transactions appear to have “stopped” before these files were created. I am not sure if this is the cause or the effect… Initially, I was under the assumption that my stops were related to PXC3706 (or maybe 3776?) which ?may? be resolved in 8.0.26-16.1 (see above post in the 8.0.26 thread)
What I noticed was a large jump in the Avg Replication delay (the delay was on the Writer I send all to one system none of the others get much traffic. and is a 5 node Cluster geographically diverse with three in one data center)
According to pmm no flow control items were started or anything else that jumps out at me…
and I saw no abnormal locks or other items… just it “stopped”
If what I am reading is correct I should be able to review the binlogs and determine the large transaction that may have caused this and ask them to set a streaming session?
**Note as I have slow query logs enabled and did not see anything at the time I am guessing that cause I had to crash it the query did not show up (As it didn’t finish) I also saw nothing in the binlog (but now I know what to look for more)
Your thoughts on if this is/was a cause or effect of the system stopping processing as writes (reads kept working)
**Note I need to enable the wait event analyses for the wsrep conditions as right now my logs appear to be empty for those graphs…
1 Like
