+1 Penghui
> On Jan 16, 2023, at 23:36, Yubiao Feng <yubiao.f...@streamnative.io.INVALID> > wrote: > > Hi community > > I am starting a DISCUSS for making the default value of the parameter > "--get-subscription-backlog-size" of admin API "topics stats" true. > > In the PR https://github.com/apache/pulsar/pull/9302, the property backlog > size of each subscription returned in the response of the API topics stats, > by default this property is always equal to 0 in response, and this will > confuse users. Since the calculation of backlog size is done in broker > memory, there is no significant overhead(the process is described in the > following section), so I think the correct values should be displayed by > default. > > ### The following two APIs should be affected: > > In Pulsar admin API > ``` > pulsar-admin topics stats persistent://test-tenant/ns1/tp1 > --get-subscription-backlog-size > pulsar-admin topics stats persistent://test-tenant/ns1/tp1 -sbs > ``` > the default value of parameter `--get-subscription-backlog-size` will be > `true` > > In Pulsar Rest API > ``` > curl GET "http://127.0.0.1:8080/test-tenant/ns1/tp1/stats > "?subscriptionBacklogSize=true > ``` > the default value of parameter `subscriptionBacklogSize ` will be `true` > > > ### The following is the process of calculating backlog size: > - Divide `PersistentTopc.ledgers` into two parts according to the ledgerId > of the mark delete position of the cursor. The second part is ledgers > indicating the messages still need to be consumed, aka backlogSizeInLedgers. > - Find the LedgerInfo whose ledgerId is the same as the ledgerId of the > mark delete position of the cursor, and we can also divide the ledger into > two parts, the second part is entries indicating the messages still need to > be consumed, multiply the average size of each entry in metrics by the > number of still need to be consumed entries we can get the backlog size in > this ledger. aka backlogSizeInEntries. > - `backlogSizeInLe dgers` + `backlogSizeInEntries` > > Thanks > Yubiao Feng
