由于记忆力低，我们的一台生产服务器每隔几周就会崩溃。我们发现实例内存是由MemoryClerk_sosnode店员拍摄的。当该店员逐渐使用可用的内存时，计划缓存变小，SQL引擎开始经常编译和重新编译计划。另外，页面预期寿命也会发生很多波动。
内存文员大小，
请单击下面的链接以查看屏幕截图。我无法将其作为图片发布，因为我在Stackoverflow中还没有足够的声誉点。
为其他内存店员分配的内存
/wycrr.jpg“ rel =“ nofollow noreferrer”>前5个内存店员大小
现在，我发现Microsoft已将此店员的修复程序逐渐在其CU11中逐渐使用，并且他们放在CU15中（不放在Cu15）当然，如果他们真的修复了）
修复了SQL 2019 Cu11
https://support.microsoft.com/en-us/topic/kb5003249-cumulative-update-11-for-sql-server-2019-657b2977-a0f1-4e1f-8b93-8c2ca8b6bef5#bkmk_13990055< /a>
修复SQL 2019 Cu15
https://support.microsoft.com/en-us/topic/kb5008996-cumulative-update-15-for-sql-server-sql-server-2019-4b6a8eee9-1c61c61-1c61-482d-914f-914f-914f-914f-36e429999999999999999999901fbb6 <
/修复说，它可能会通过使用空间索引引起。我们根本不使用它。由于微软有KB文章说他们将其修复在Cu11中，然后在CU15中再次修复。我很确定，这里有问题。
谁能在这里帮助我缩小问题？
谢谢。
One of our production server is crashing after every few weeks due to low memory. We found that the instance memory is taken by MEMORYCLERK_SOSNODE clerk. When this clerk gradually use available memory, the plan cache becomes small and SQL engine starts compiling and re-compiling plans frequently. Also, the page life expectancy fluctuates a lot.
Memory clerk size
Please click below links to see the screenshot. I'm not able to post it as a picture here as I don't have enough reputation points in stackoverflow yet.
Memory assigned for other memory clerk
top 5 memory clerk size
Now, I found that the Microsoft has put fix for this clerk gradually taking available memory in their CU11 and same fix they put in CU15 (Not sure if they actually fixed it)
Fix in SQL 2019 CU11
https://support.microsoft.com/en-us/topic/kb5003249-cumulative-update-11-for-sql-server-2019-657b2977-a0f1-4e1f-8b93-8c2ca8b6bef5#bkmk_13990055
Fix in SQL 2019 CU15
https://support.microsoft.com/en-us/topic/kb5008996-cumulative-update-15-for-sql-server-2019-4b6a8ee9-1c61-482d-914f-36e429901fb6
Both fix says, it can cause by using spatial index. We are not using it at all. Since Microsoft has KB article says they fixed it in CU11 and then again in CU15. I'm pretty sure, something is wrong here.
Can anyone help me narrow down the problem here?
Thank you.
发布评论
评论（1）
根据也有一个修复记忆泄漏的修复程序：
2068768-“修复了由日志读取器代理在交易复制中引起的SQL Server进程（MemoryClerk_sosnode下的高用法）中的逐渐内存泄漏。”
According to https://learn.microsoft.com/en-us/troubleshoot/sql/releases/sqlserver-2019/cumulativeupdate19 there is also a fix for the memory leak:
2068768 - "Fixes a gradual memory leak in the SQL Server process (the high usage under MEMORYCLERK_SOSNODE) caused by the Log Reader Agent in transactional replication."