Details
-
Bug
-
Resolution: Fixed
-
Major
-
7.1.0
-
Triaged
-
n/a
-
1
-
No
Description
Immediately after launching a 7.1.0 node on Amazon Linux 2 - Graviton instance, systemStats on /pools/nodes reports an abnormally large number (greater than UInt64).
"systemStats": { |
"allocstall": 18446744073709552000 |
},
|
Whereas non-arm builds report allocstall as zero
"systemStats": { |
"allocstall": 0 |
}
|
After some digging, we have discovered that the issue arrises from an underflow in sigar. Here there is an if statement that falls into the else because on ARM, the /proc/vmstat does not return a allocstall, or allocstall_dma32
[ec2-user@ip-172-31-7-174 ~]$ cat /proc/vmstat | grep allocstall |
allocstall_dma 0 |
allocstall_normal 0 |
allocstall_movable 0 |
This causes a -1 to be assigned to the reply.allocstall field which is a uint64_t which is defined here