Open edolstra opened 8 years ago
Queue runner status after 76,000s of uptime:
"s3" : {
"costDollarApprox" : 12.7399,
"get" : 346118,
"getBytes" : 143014408725,
"getSpeed" : 3.77933,
"getTimeMs" : 36088182,
"head" : 117607,
"put" : 113421,
"putBytes" : 389521856099,
"putSpeed" : 2.44061,
"putTimeMs" : 152206399
}
So this is fixed?
Need to figure out why this happens:
The first download is probably due to caching of negative lookups in Cloudfront, causing a build machine to be unable to fetch the path directly. This can be fixed by having the build machines access the underlying S3 bucket directly, rather than Cloudfront. (This would also save some money for EC2 build machines.)
The second download is probably due to the queue monitor thread encountering a new build of a cached path, requiring it to fetch the path to determine the build products. This could probably be avoided by copying the build products from the previous build in the Hydra database.