In addition, We can redefine the retry strategy using retries(0..Integer.MAX_VALUE, default is 0) combined retry interval(1s..Ns, default is 1s)
Never retry, retries=0 Retry N, retries=N (default duration is N * 1s) Retry until succeed(Really? Maybe we just want to try within a certain time limit), E.g. retries=3600 and retryInterval=1s (which means retry per second within an hour) Best regards, Lei Zhang On December 12, 2019 at 10:35:47 PM, Willem Jiang (willem.ji...@gmail.com) wrote: Maybe we can consider to disable the forward recovering with boolean flag instead of setting the retries to 0. Willem Jiang Twitter: willemjiang Weibo: 姜宁willem On Thu, Dec 12, 2019 at 3:03 PM Zhang Lei <coolbee...@gmail.com> wrote: > > Hi, Feng Zheng > > Thank you for your reply. > > The question stems from the GitHub issue[1] > > I haven't paid attention to the forward recovery before. When I optimized > the state machine compensation [2] recently, I found that I used retry and > timeout incorrectly in the compensation strategy. For this, I sent an > email[3] discussing it. > > To distinguish between forward and reverse recovery properties, I have > created a JIRA [4] to track it. I plan to add the properties reverseRetries > and reverseTimeout to @Compensable in another PR and use it in the state > machine compensation strategy > > About the combination of forward and reverse compensation, I will create a > new JIRA to describe it, but I think this may cause some modification of > the forward recovery test case. > > [ <https://github.com/apache/servicecomb-pack/pull/609>1] > https://github.com/apache/servicecomb-pack/issues/589 > [2] https://issues.apache.org/jira/browse/SCB-1622 > [3] > https://lists.apache.org/thread.html/a81f042ba5cc711c0df0a1fe7144b5ed6e9f20336534ccf67d77dcf1%40%3Cdev.servicecomb.apache.org%3E > [4] https://issues.apache.org/jira/browse/SCB-1646 > > > Best regards, > Lei Zhang > > On December 12, 2019 at 12:32:32 PM, Zheng Feng (zh.f...@gmail.com) wrote: > > So you want to introduce a new policy for recovering ? > > "try forward recovering at first in retries times, if it fails then > following with the reverse to compensate." This makes sense but this could > add a new attribute with the @Compensable. > I'm not sure if it could be incompatible. > > Anyway, try to raise a JIRA and record this idea.
