Hi all,
In the IoTDB test, we find a memory leak bug due to the memory controller module. As test goes by, more and more metadata objects are cached by the memory controller and cannot be released, leading to frequently GC and the process crash. The developer Jian Tian has proposed a PR to solve it: https://github.com/apache/incubator-iotdb/pull/149 We hope to fix this bug before the first version release to ensure the stability of IoTDB. — 顺颂时祺 康荣 清华大学软件学院 — Best Regards, Rong Kang School of Software, Tsinghua University 原始邮件 发件人:Xiangdong huangsaint...@gmail.com 收件人:dev...@iotdb.apache.org 发送时间:2019年4月18日(周四) 21:42 主题:Re: For the first release version Hi all, Thanks for Christofer's reminder. “We will work on getting this set up with in a week or so. Please provide a list of the apache IDs of the people who will be responsible for installing and maintaining the software on the VM. ” If someone want to have the privilege of controlling the VM, please comment under this mail thread. Firstly, I apply for that. Best, ----------------------------------- Xiangdong Huang School of Software, Tsinghua University 黄向东 清华大学 软件学院 Christofer Dutz christofer.d...@c-ware.de 于2019年4月18日周四 下午8:24写道： Hu Guys, infra is asking for the apache ids of people needing access to the machine. Chris Am 12.04.19, 09:08 schrieb "Xiangdong Huang" saint...@gmail.com: Hi, I have created a ticket on JIRA [1] for getting a VM. :D [1] https://issues.apache.org/jira/browse/INFRA-18202 ----------------------------------- Xiangdong Huang School of Software, Tsinghua University 黄向东 清华大学 软件学院 Christofer Dutz christofer.d...@c-ware.de 于2019年4月10日周三 下午4:27写道： Hi Xiangdong, having an Apache VM has the additional benefit of having everything at Apache. I have been involved in other Apache projects where external VMs were involved. Things become problematic if the key resource for accessing this is unavailable (Leaves the project, is on holidays, is sick, ...). I guess it wouldn't really matter when you do the split, ideally it should be when the Master branch is in a usable state ... then after your first release, you'll merge the Release version back to master and from there on continue to only merge releases There. The Jenkinsfile is already prepared for this setup (it sends warning emails for commits To master, but that's currently commented out). So prior to doing the split, give me a short signal and I'll prepare the Jenkinsfile for that branching mode. Chris Am 10.04.19, 10:19 schrieb "Xiangdong Huang" saint...@gmail.com: Hi, OK, I will try to apply a VM from INFRA. When I say a long-term test, I hope a test at least one day... Now I am planning to run two types of tests in our lib: - A long term test that lasts for either one week, or the main dev branch is updated. - A aging test that runs all the time until the IoTDB instance crashes or there is no disk space... If we have a VM, at least we can run a long-term test that lasts for one day, which will trigger many times of the following tasks: flushing to disk, closing TsFiles, merging TsFiles with Overflow files. It is helpful for guaranteeing IoTDB is stable. This morning I see the issue is solved by PR [1]. The thing that we need to discuss is, which source code version (commitlog id) do we set the master branch to? Current commitlog id? or splitting the functionality of master/dev until we release the first version? (I remember on the webpage that Julian shared, the master branch should stay on a tagged version, e.g., 0.1, 0.2, etc..) [1] https://github.com/apache/incubator-iotdb/pull/138 Best, ----------------------------------- Xiangdong Huang School of Software, Tsinghua University 黄向东 清华大学 软件学院 Julian Feinauer j.feina...@pragmaticminds.de 于2019年4月10日周三 下午3:22写道： Also, +1 from my side for dev / master. And indeed, a separate VM for this kind of tests would be a good idea. And we could either manage that via separate Jenkins jobs or manual before releases or so. How long do you mean with long-term? Hours, Days, Weeks, ... ? Julian Am 10.04.19, 09:19 schrieb "Christofer Dutz" christofer.d...@c-ware.de: Hi Xiangdong, you can ask for a VM from Infra, if this is needed. We did that for the PLC4X project as we needed to do network stuff infra didn't want to have on it's official nodes. And I agree having master and develop branches is a very good thing. Chris Am 10.04.19, 05:12 schrieb "Xiangdong Huang" saint...@gmail.com : Hi, Sadly, a long term test failed, see the JIRA for more details [1]. The problem is caused by the modification file management, which is introduced in PR [2]. Firstly, we need to fix this bug ASAP, otherwise the master branch does not work...... ( @Tian Jiang, the author of PR [2]). And then we can keep go on for the release version. Secondly, it shows that for a DB, just a UT/IT test with JUnit is not enough. Instead, a long term test is always needed once the main branch of code is updated... It seems that there is no such a public platform can support long term test. I think we can open our lib's test environment for IoTDB if needed. Thirdly, it is time to enable the dev branch... (I remember that Julian shared a link to introduce the functionality of the mater branch and dev branch). We need to keep the master branch is stable all the time, and develop in the dev branch. If we enable that, we need to quickly resolve current bug to make the master branch stable, and moving all PRs to the dev branch, rather than the master branch.. Best, [1] https://issues.apache.org/jira/projects/IOTDB/issues/IOTDB-79 [2] https://github.com/apache/incubator-iotdb/pull/17 ----------------------------------- Xiangdong Huang School of Software, Tsinghua University 黄向东 清华大学 软件学院 Xiangdong Huang saint...@gmail.com 于2019年4月3日周三 上午10:31写道： Hi, Now: - The Polding name search has finished [1] - The license self-check is finished [2] And we have some tasks left: - (1) prepare the final code version. - (2) modify the changes file (we have an initial version) - (3) begin to vote... For task (1) and (2): The following PRs have opened for several days, if there is no other feedback, I will merge them: #79 [IOTDB-6]Value filter query optimization #97 [IOTDB-47][IOTDB-54][IOTDB-59][IOTDB-60]Aggregate+GroupBy+Fill #111 try to release memory asap in ReadOnlyMemChunk #123 provide unified query resource control interface Then I will start a long-term test (at least 5 days) with a heavy write workload for checking whether the current version is stable.... If everything is fine, we can begin task (2) and add a tag on the repository and finish task (1) and (2). And then, begin (3)... [1] https://lists.apache.org/thread.html/9d26c2eab119d345a542edd5bc5a657d451fbd1a92b33bad1fd74af5@%3Cdev.iotdb.apache.org%3E [2] https://lists.apache.org/thread.html/2bb4abd5a1cf9961dfbe50182bcf7cb39fcd408baa5b90fe4ddb0bd3@%3Cdev.iotdb.apache.org%3E Best, ----------------------------------- Xiangdong Huang School of Software, Tsinghua University 黄向东 清华大学 软件学院
