FIND NOLOOP PATH WITH PROP FROM 导致内存爆掉

问答 问题
,
1

当前版本v3.0，
当前库中7k 节点，3w条边
当前主机8c，64GB
1~3月份这条语句可以执行出来（只是慢点），到了4月份（导致内爆炸），其间数据规模没有增加，设备没有变动，这个是为什么，有什么改进方式？【1、当前查询语法适配我们的业务，不能改；2、 10 STEPS 语句是可以执行的】

FIND NOLOOP PATH WITH PROP FROM 100422 TO 100324 OVER edge_join where edge_join.road_fid not in [3835,3921] and edge_join._src in [1917,…,100344,100614] and edge_join._dst in [1917,1974,1961,1919,…1736,100344,100614] and (edge_join.edge_type in [“1”] or ((edge_join._src in [100422,100324] or edge_join._dst in [100422,100324 ]) and edge_join.edge_type in [“2”] ) ) UPTO 14 STEPS YIELD path as p |limit 100 ;

2

来按照这个帖子，把相关的信息补充下。

3

机器数 :1
磁盘类型 ：HDD
CPU 核数 :4C
内存大小:64GB
NebulaGraph 版本号 3.0

服务状态，使用 show hosts 语句

image
image1510×345 24.5 KB

图空间 : optical_route_plan_ZH

查询语句，需要调优的那条语句

FIND NOLOOP PATH WITH PROP FROM 100422 TO 100324  OVER edge_join where edge_join.road_fid not in [3835,3921] and edge_join._src in [1917,1974,1961,1919,101153,1970,1988,1951,1667,1661,1810,1811,1800,100658,1815,1807,1834,1472,1468,1777,1776,1802,1801,1857,1692,1695,1831,1775,1792,1669,1675,1605,1688,1644,1604,1634,1665,101252,1673,1703,1799,1827,1728,1727,1804,1805,1816,1733,1895,1751,1761,1778,100229,100230,1793,1809,1824,1700,1852,1720,1737,1749,1770,1790,1818,1765,1704,1830,1740,1806,1759,100063,1902,1791,1890,1907,1698,1814,1697,1682,1649,1668,1640,1678,1503,1511,100032,100031,1500,1721,1750,1783,1803,100062,1742,100517,1773,1694,100423,100422,100334,1734,100335,1766,1781,1730,1696,100324,1722,1508,1613,1620,1663,1674,1618,1619,1625,1581,1639,1861,1756,1710,1747,1707,1764,1744,1702,1882,1741,1656,1670,1626,1565,1652,1657,1642,1629,1680,1496,1828,1699,1739,1752,1872,1871,1746,1715,1723,1731,1647,1662,1631,100276,1654,1651,1638,1587,1645,1540,1817,1849,100439,100440,1859,1839,1889,1822,1691,1838,100277,1637,1684,1588,1623,1528,100896,1672,1612,1608,1607,1610,1611,1602,100520,1666,1628,1595,1650,1876,1846,1758,1788,1795,1797,1796,1853,1789,1701,1636,1614,1627,100975,1709,1714,1736,100344,100614] and edge_join._dst in [1917,1974,1961,1919,101153,1970,1988,1951,1667,1661,1810,1811,1800,100658,1815,1807,1834,1472,1468,1777,1776,1802,1801,1857,1692,1695,1831,1775,1792,1669,1675,1605,1688,1644,1604,1634,1665,101252,1673,1703,1799,1827,1728,1727,1804,1805,1816,1733,1895,1751,1761,1778,100229,100230,1793,1809,1824,1700,1852,1720,1737,1749,1770,1790,1818,1765,1704,1830,1740,1806,1759,100063,1902,1791,1890,1907,1698,1814,1697,1682,1649,1668,1640,1678,1503,1511,100032,100031,1500,1721,1750,1783,1803,100062,1742,100517,1773,1694,100423,100422,100334,1734,100335,1766,1781,1730,1696,100324,1722,1508,1613,1620,1663,1674,1618,1619,1625,1581,1639,1861,1756,1710,1747,1707,1764,1744,1702,1882,1741,1656,1670,1626,1565,1652,1657,1642,1629,1680,1496,1828,1699,1739,1752,1872,1871,1746,1715,1723,1731,1647,1662,1631,100276,1654,1651,1638,1587,1645,1540,1817,1849,100439,100440,1859,1839,1889,1822,1691,1838,100277,1637,1684,1588,1623,1528,100896,1672,1612,1608,1607,1610,1611,1602,100520,1666,1628,1595,1650,1876,1846,1758,1788,1795,1797,1796,1853,1789,1701,1636,1614,1627,100975,1709,1714,1736,100344,100614] and (edge_join.edge_type in ["1"] or ((edge_join._src  in [100422,100324] or  edge_join._dst  in [100422,100324 ]) and edge_join.edge_type in ["2"] ) )   UPTO 14 STEPS  YIELD path as p  |limit 100 ;

数据量，7k 节点，3w条边

profile 信息，使用 profile query 语句

4

result1.csv (45.9 KB)

5

你数据有没有新增删除？

6

nebula graph升级了么

7

没有

9

能升级一下版本吗，因为 find path 在最新的版本上做了一些优化，内存和性能都有些提升

10

此话题已在最后回复的 30 天后被自动关闭。不再允许新回复。