Description
We have seen issues that capi replication restarted repeatedly. This is especially problematic in backfill mode where replication, on restart, may rollback to 0 and hence not making progress. We need to make a number of capi runtime parameters configurable to make it possible to reduce the frequency of capi replication restart.
Following is the list of such parameters and their default values:
// max retry for capi batchUpdateDocs operation
var CapiMaxRetryBatchUpdateDocs = 6
// timeout for batch processing in capi
// 1. http timeout in revs_diff, i.e., batchGetMeta, call to target
// 2. overall timeout for batchUpdateDocs operation
var CapiBatchTimeout = 180 * time.Second
// timeout for tcp write operation in capi
var CapiWriteTimeout = 10 * time.Second
// timeout for tcp read operation in capi
var CapiReadTimeout = 60 * time.Second
Their values can be changed at runtime through the following non-public rest api:
curl -X POST -u Administrator:welcome http://[hostname]:9998/xdcr/internalSettings -d CapiMaxRetryBatchUpdateDocs=10
Attachments
Issue Links
For Gerrit Dashboard: MB-23305 | ||||||
---|---|---|---|---|---|---|
# | Subject | Branch | Project | Status | CR | V |
75088,7 | MB-23305 make capi runtime parameters configurable | master | goxdcr | Status: MERGED | +2 | +1 |
75801,2 | MB-23305 make capi runtime parameters configurable | watson | goxdcr | Status: MERGED | +2 | +1 |
75813,4 | MB-23305 use write timeout for write to part ch timeout | master | goxdcr | Status: MERGED | +2 | +1 |
75827,2 | MB-23305 use write timeout for write to part ch timeout | watson | goxdcr | Status: MERGED | +2 | +1 |
75837,3 | MB-23305 make capi runtime parameters configurable | 4.6.1 | goxdcr | Status: MERGED | +2 | +1 |
75840,2 | MB-23305 use write timeout for write to part ch timeout | 4.6.1 | goxdcr | Status: MERGED | +2 | +1 |