The invention discloses an erasure code filing method based on data cache in a cluster. The method comprises following steps of (1), according to a user access request, reading demanded data subblocks from production nodes in which the data subblocks are located to a node buffer; sending the data subblocks to an access node; forwarding the data subblocks to a filing node according to a filing data state table; (2), updating the number of the data subblocks of each data node by the filing node according to the received data subblocks; judging whether the filing node receives all data subblocks of current strips; if so, filing the current strips; otherwise, reading cold data subblocks from a production cluster, filing after the strips are gathered completely. In adoption of the erasure code filing method provided by the invention, compared with the normal erasure code filing method, because the user access cached data subblocks are adopted, the number of times of reading the data subblocks from the production cluster by the filing node is greatly reduced; the erasure code filing operation is carried out under the condition of not influencing the user access; and the filing efficiency is improved.