时间戳

首页 > 解决方案 > H2O 减少每个预测的可用内存 - 结果 OOM

python-3.x - H2O 减少每个预测的可用内存 - 结果 OOM

问题描述

我用 h2o 构建应用程序。每次调用我的应用程序都会对 H2O 进行预测。H2O 瞬间从应用程序开始，然后每个呼叫都连接到这个瞬间。

我的问题是，随着时间的推移，我的可用内存计数器正在减少。

这是我在连接到现有 H2O 实例期间的日志。

2020-04-08 12:11:23.821 --------------------------  ---------------------------------------------------
2020-04-08 12:11:23.821 Python version:             3.6.4 final
2020-04-08 12:11:23.821 H2O API Extensions:         Amazon S3, XGBoost, Algos, AutoML, Core V3, Core V4
2020-04-08 12:11:23.821 H2O internal security:      False
2020-04-08 12:11:23.821 H2O connection proxy:
2020-04-08 12:11:23.821 H2O connection url:         http://localhost:54321
2020-04-08 12:11:23.821 H2O cluster status:         locked, healthy
2020-04-08 12:11:23.821 H2O cluster allowed cores:  2
2020-04-08 12:11:23.821 H2O cluster total cores:    2
2020-04-08 12:11:23.821 H2O cluster free memory:    5.903 Gb
2020-04-08 12:11:23.821 H2O cluster total nodes:    1
2020-04-08 12:11:23.821 H2O cluster name:           H2O_from_python_cdsw_dwwqwt
2020-04-08 12:11:23.821 H2O cluster version age:    1 year and 7 days !!!
2020-04-08 12:11:23.821 H2O cluster version:        3.24.0.1
2020-04-08 12:11:23.821 H2O data parsing timezone:  UTC
2020-04-08 12:11:23.821 H2O cluster timezone:       Europe/Zurich
2020-04-08 12:11:23.821 H2O cluster uptime:         39 mins 45 secs
2020-04-08 12:11:23.821 --------------------------  ---------------------------------------------------

2020-04-08 12:11:46.578 --------------------------  ---------------------------------------------------
2020-04-08 12:11:46.578 Python version:             3.6.4 final
2020-04-08 12:11:46.578 H2O API Extensions:         Amazon S3, XGBoost, Algos, AutoML, Core V3, Core V4
2020-04-08 12:11:46.578 H2O internal security:      False
2020-04-08 12:11:46.578 H2O connection proxy:
2020-04-08 12:11:46.578 H2O connection url:         http://localhost:54321
2020-04-08 12:11:46.578 H2O cluster status:         locked, healthy
2020-04-08 12:11:46.578 H2O cluster allowed cores:  2
2020-04-08 12:11:46.578 H2O cluster total cores:    2
2020-04-08 12:11:46.577 H2O cluster free memory:    5.890 Gb
2020-04-08 12:11:46.577 H2O cluster total nodes:    1
2020-04-08 12:11:46.577 H2O cluster name:           H2O_from_python_cdsw_v1eq0y
2020-04-08 12:11:46.577 H2O cluster version age:    1 year and 7 days !!!
2020-04-08 12:11:46.577 H2O cluster version:        3.24.0.1
2020-04-08 12:11:46.577 H2O data parsing timezone:  UTC
2020-04-08 12:11:46.577 H2O cluster timezone:       Europe/Zurich
2020-04-08 12:11:46.577 H2O cluster uptime:         40 mins 8 secs
2020-04-08 12:11:46.577 --------------------------  ---------------------------------------------------

它一直工作直到它有内存然后它停止工作（因为OOM）。我该如何处理这种情况？我可以按计划删除某些内容（旧日志）以保持可用内存水平不变吗？

标签： python-3.xmemoryh2o

解决方案

推荐阅读